Inizia

Guida introduttiva: NVIDIA Jetson con Ultralytics YOLO26

Questa guida illustra la distribuzione di Ultralytics YOLO26 sui dispositivi NVIDIA Jetson e include benchmark delle prestazioni che mostrano le capacità di YOLO26 su questi dispositivi compatti ed efficienti dal punto di vista energetico.

Supporto per i nuovi prodotti

Abbiamo aggiornato questa guida con il più recente NVIDIA Jetson AGX Thor Developer Kit, che offre fino a 2070 TFLOPS FP4 di potenza di calcolo AI e 128 GB di memoria, con potenza configurabile tra 40 W e 130 W. Offre una potenza di calcolo AI oltre 7,5 volte superiore a NVIDIA Jetson AGX Orin, con un'efficienza energetica 3,5 volte maggiore nell'esecuzione dei modelli AI più diffusi.



Guarda: Sblocca un'inferenza più veloce con Ultralytics YOLO26 sui dispositivi NVIDIA Jetson 🚀
NVIDIA Jetson Ecosystem
Nota

Questa guida è stata testata con NVIDIA Jetson AGX Thor Developer Kit (Jetson T5000) e NVIDIA Jetson AGX Orin Developer Kit (64GB), con l'ultima versione stabile di JetPack 7.2; con NVIDIA Jetson Orin Nano Super Developer Kit, con la versione JP6.1 di JetPack; con Seeed Studio reComputer J4012, basato su NVIDIA Jetson Orin NX 16GB e con la versione JP6.0/JP5.1.3 di JetPack; e con Seeed Studio reComputer J1020 v2, basato su NVIDIA Jetson Nano 4GB e con la versione JP4.6.1 di JetPack. È previsto che funzioni con l'intera gamma di hardware NVIDIA Jetson, compresi i dispositivi più recenti e quelli meno recenti.

Che cos'è NVIDIA Jetson?#

NVIDIA Jetson è una serie di schede di calcolo embedded progettate per portare l'elaborazione accelerata dell'intelligenza artificiale (AI) sui dispositivi edge. Questi dispositivi compatti e potenti si basano sull'architettura GPU di NVIDIA e possono eseguire direttamente sul dispositivo algoritmi complessi di intelligenza artificiale e modelli di apprendimento profondo, senza dipendere da risorse di elaborazione cloud. Le schede Jetson sono spesso utilizzate nella robotica, nei veicoli autonomi, nell'automazione industriale e in altre applicazioni in cui l'inferenza AI deve essere eseguita localmente con bassa latenza ed elevata efficienza. Inoltre, queste schede si basano sull'architettura ARM64 e consumano meno energia rispetto ai tradizionali dispositivi di calcolo GPU.

Confronto tra le serie NVIDIA Jetson#

NVIDIA Jetson AGX Thor è l'ultima versione della famiglia NVIDIA Jetson, basata sull'architettura NVIDIA Blackwell, che offre prestazioni AI nettamente superiori rispetto alle generazioni precedenti. La tabella seguente confronta alcuni dispositivi Jetson dell'ecosistema.

Jetson AGX Thor (T5000)Jetson AGX Orin 64GBJetson Orin NX 16GBJetson Orin Nano SuperJetson AGX XavierJetson Xavier NXJetson Nano
Prestazioni AI2070 TFLOPS275 TOPS100 TOPS67 TOPS32 TOPS21 TOPS472 GFLOPS
GPUGPU con architettura NVIDIA Blackwell a 2560 core e 96 Tensor CoreGPU con architettura NVIDIA Ampere a 2048 core e 64 Tensor CoreGPU con architettura NVIDIA Ampere a 1024 core e 32 Tensor CoreGPU con architettura NVIDIA Ampere a 1024 core e 32 Tensor CoreGPU con architettura NVIDIA Volta a 512 core e 64 Tensor CoreGPU con architettura NVIDIA Volta™ a 384 core e 48 Tensor CoreGPU con architettura NVIDIA Maxwell™ a 128 core
Frequenza massima della GPU1.57 GHz1.3 GHz918 MHz1020 MHz1377 MHz1100 MHz921MHz
CPUCPU Arm® Neoverse®-V3AE a 64 bit e 14 core, 1 MB L2 + 16 MB L3CPU NVIDIA Arm® Cortex A78AE v8.2 a 64 bit e 12 core, 3 MB L2 + 6 MB L3CPU NVIDIA Arm® Cortex A78AE v8.2 a 64 bit e 8 core, 2 MB L2 + 4 MB L3CPU Arm® Cortex®-A78AE v8.2 a 64 bit e 6 core, 1,5 MB L2 + 4 MB L3CPU NVIDIA Carmel Arm®v8.2 a 64 bit e 8 core, 8 MB L2 + 4 MB L3CPU NVIDIA Carmel Arm®v8.2 a 64 bit e 6 core, 6 MB L2 + 4 MB L3Processore quad-core Arm® Cortex®-A57 MPCore
Frequenza massima della CPU2.6 GHz2.2 GHz2.0 GHz1.7 GHz2.2 GHz1.9 GHz1.43GHz
Memoria128 GB 256-bit LPDDR5X 273 GB/s64 GB 256-bit LPDDR5 204,8 GB/s16 GB 128-bit LPDDR5 102,4 GB/s8 GB 128-bit LPDDR5 102 GB/s32 GB 256-bit LPDDR4x 136,5 GB/s8 GB 128-bit LPDDR4x 59,7 GB/s4 GB 64-bit LPDDR4 25,6 GB/s

Che cos'è NVIDIA JetPack?#

NVIDIA JetPack SDK alimenta i moduli Jetson e fornisce un ambiente di sviluppo completo per creare applicazioni AI accelerate end-to-end. JetPack include Jetson Linux con bootloader, kernel Linux, ambiente desktop Ubuntu e una suite completa di librerie per accelerare il calcolo GPU, i contenuti multimediali, la grafica e la visione artificiale. Include anche esempi, documentazione e strumenti per sviluppatori sia per il computer host sia per il kit di sviluppo, e supporta SDK di livello superiore come DeepStream per l'analisi video in streaming, Isaac per la robotica e Riva per l'AI conversazionale.

Installa JetPack su NVIDIA Jetson#

Il primo passaggio dopo aver ricevuto un dispositivo NVIDIA Jetson è installare NVIDIA JetPack sul dispositivo. Esistono diversi metodi per installare NVIDIA JetPack sui dispositivi NVIDIA Jetson.

  1. Per JetPack 7.2 su un Jetson AGX Thor, AGX Orin o Orin Nano Developer Kit ufficiale, scarica l'ISO unificata Jetson, scrivila su un'unità flash USB e segui la guida introduttiva specifica per AGX Thor, AGX Orin o Orin Nano. A partire da JetPack 7.2, Orin Nano non utilizza più un'immagine della scheda SD scaricabile; l'ISO su USB installa Jetson Linux sulla scheda microSD o sull'SSD NVMe del dispositivo.
  2. Se vuoi utilizzare JetPack 6 su un Jetson Orin Nano Developer Kit, segui le istruzioni di NVIDIA per l'aggiornamento a JetPack 6.x e l'uso della scheda SD.
  3. Se possiedi un altro Development Kit NVIDIA, puoi installare JetPack sul dispositivo con SDK Manager.
  4. Se possiedi un dispositivo Seeed Studio reComputer J4012, puoi installare JetPack sull'SSD incluso; se invece possiedi un dispositivo Seeed Studio reComputer J1020 v2, puoi installare JetPack sull'eMMC/ SSD.
  5. Se possiedi un altro dispositivo di terze parti basato sul modulo NVIDIA Jetson, ti consigliamo di seguire la procedura di installazione tramite riga di comando.
Nota

Per i metodi 1, 4 e 5 riportati sopra, dopo aver eseguito il flashing del sistema e avviato il dispositivo, esegui sudo apt update && sudo apt install nvidia-jetpack -y nel terminale del dispositivo per installare tutti i componenti JetPack rimanenti necessari.

Supporto di JetPack in base al dispositivo Jetson#

La tabella seguente mostra le versioni di NVIDIA JetPack supportate dai diversi dispositivi NVIDIA Jetson.

JetPack 4JetPack 5JetPack 6JetPack 7
Jetson Nano✅❌❌❌
Jetson TX2✅❌❌❌
Jetson Xavier NX✅✅❌❌
Jetson AGX Xavier✅✅❌❌
Jetson AGX Orin❌✅✅✅
Jetson Orin NX❌✅✅✅
Jetson Orin Nano❌✅✅✅
Jetson AGX Thor❌❌❌✅

Guida introduttiva a Docker#

Il modo più rapido per iniziare a usare Ultralytics YOLO26 su NVIDIA Jetson è eseguire le immagini Docker precompilate per Jetson. Consulta la tabella sopra e scegli la versione di JetPack in base al dispositivo Jetson che possiedi.

t=ultralytics/ultralytics:latest-jetson-jetpack4
sudo docker pull $t && sudo docker run -it --ipc=host --runtime=nvidia $t
Supporto di Docker e TensorRT per JetPack 7

L'immagine latest-nvidia-arm64 usa NVIDIA PyTorch 26.10, che include CUDA 13.4 e TensorRT 11.4. NVIDIA indica JetPack 7.1 per PyTorch nella propria tabella di compatibilità, ma TensorRT 11 non supporta JetPack, Thor incluso. Usa questa immagine solo per carichi di lavoro PyTorch su un host supportato. Per le esportazioni TensorRT su Jetson, usa l'installazione nativa riportata di seguito con il runtime TensorRT 10.x supportato dalla tua versione di JetPack. JetPack 7.2 su Thor o Orin richiede una convalida separata del container. Ricrea i motori per la GPU di destinazione e la versione di TensorRT.

Per le immagini JetPack 4, 5 e 6, continua con Usa TensorRT su NVIDIA Jetson. L'immagine JetPack 7.1 qui sopra è solo per i carichi di lavoro PyTorch.

Inizia con l'installazione nativa#

Per un'installazione nativa senza Docker, segui i passaggi riportati di seguito.

Esegui su JetPack 7.2#

Installa il pacchetto Ultralytics#

Installa il pacchetto Ultralytics su Jetson. Le dipendenze per l'esportazione vengono installate automaticamente la prima volta che esporti un modello, quindi qui basta il pacchetto di base. Questa guida si concentra sulle esportazioni NVIDIA TensorRT, perché TensorRT offre le migliori prestazioni sui dispositivi Jetson.

  1. Aggiorna l'elenco dei pacchetti, installa pip e aggiorna pip all'ultima versione

    sudo apt update
    sudo apt install python3-pip -y
    pip install -U pip
  2. Installa il pacchetto pip ultralytics

    pip install ultralytics
  3. Riavvia il dispositivo

    sudo reboot

Installa PyTorch e Torchvision#

L'installazione di ultralytics descritta sopra installerà Torch e Torchvision. Tuttavia, questi due pacchetti installati tramite pip non sono compatibili con i dispositivi JetPack 7.2 che eseguono CUDA 13, quindi devono essere installati manualmente.

Installa torch e torchvision in base a JP7.2

pip install torch torchvision --index-url https://download.pytorch.org/whl/cu130

Installa onnxruntime-gpu#

Usa un wheel GPU di ONNX Runtime compatibile con le versioni di JetPack, Python e CUDA. Le versioni attuali di PyPI includono wheel ARM64, ma non sostituiscono i pacchetti specifici del dispositivo per ogni versione di JetPack.

Scarica e installa onnxruntime-gpu 1.24.0 con il supporto Python3.12.

pip install https://github.com/ultralytics/assets/releases/download/v0.0.0/onnxruntime_gpu-1.24.0-cp312-cp312-linux_aarch64.whl

Esegui su JetPack 6.1#

Installa il pacchetto Ultralytics#

Installa il pacchetto Ultralytics su Jetson. Le dipendenze per l'esportazione vengono installate automaticamente la prima volta che esporti un modello, quindi qui basta il pacchetto di base. Questa guida si concentra sulle esportazioni NVIDIA TensorRT, perché TensorRT offre le migliori prestazioni sui dispositivi Jetson.

  1. Aggiorna l'elenco dei pacchetti, installa pip e aggiorna pip all'ultima versione

    sudo apt update
    sudo apt install python3-pip -y
    pip install -U pip
  2. Installa il pacchetto pip ultralytics

    pip install ultralytics
  3. Riavvia il dispositivo

    sudo reboot

Installa PyTorch e Torchvision#

L'installazione di ultralytics descritta sopra installerà Torch e Torchvision. Tuttavia, questi due pacchetti installati tramite pip non sono compatibili con la piattaforma Jetson, basata sull'architettura ARM64, quindi devi installare manualmente un wheel PyTorch precompilato per pip e compilare o installare Torchvision dai sorgenti.

Installa torch 2.10.0 e torchvision 0.25.0 in base a JP6.1

pip install https://github.com/ultralytics/assets/releases/download/v0.0.0/torch-2.10.0-cp310-cp310-linux_aarch64.whl
pip install https://github.com/ultralytics/assets/releases/download/v0.0.0/torchvision-0.25.0-cp310-cp310-linux_aarch64.whl
Nota

Visita la pagina PyTorch per Jetson per accedere a tutte le diverse versioni di PyTorch per le diverse versioni di JetPack. Per un elenco più dettagliato della compatibilità tra PyTorch e Torchvision, visita la pagina sulla compatibilità di PyTorch e Torchvision.

Installa cuDSS per risolvere un problema di dipendenza con torch 2.10.0

wget https://developer.download.nvidia.com/compute/cudss/0.7.1/local_installers/cudss-local-tegra-repo-ubuntu2204-0.7.1_0.7.1-1_arm64.deb
sudo dpkg -i cudss-local-tegra-repo-ubuntu2204-0.7.1_0.7.1-1_arm64.deb
sudo cp /var/cudss-local-tegra-repo-ubuntu2204-0.7.1/cudss-*-keyring.gpg /usr/share/keyrings/
sudo apt-get update
sudo apt-get -y install cudss

Installa onnxruntime-gpu#

Usa un wheel GPU di ONNX Runtime compatibile con le versioni di JetPack, Python e CUDA. Le versioni attuali di PyPI includono wheel ARM64, ma non sostituiscono i pacchetti specifici del dispositivo per ogni versione di JetPack.

Puoi trovare tutti i pacchetti onnxruntime-gpu disponibili, organizzati per versione di JetPack, versione di Python e altri dettagli di compatibilità, nella matrice di compatibilità di Jetson Zoo ONNX Runtime.

Per JetPack 6 con supporto Python 3.10, puoi installare onnxruntime-gpu 1.23.0:

pip install https://github.com/ultralytics/assets/releases/download/v0.0.0/onnxruntime_gpu-1.23.0-cp310-cp310-linux_aarch64.whl

In alternativa, per onnxruntime-gpu 1.20.0:

pip install https://github.com/ultralytics/assets/releases/download/v0.0.0/onnxruntime_gpu-1.20.0-cp310-cp310-linux_aarch64.whl

Esegui su JetPack 5.1.2#

Installa il pacchetto Ultralytics#

Installa il pacchetto Ultralytics su Jetson. Le dipendenze per l'esportazione vengono installate automaticamente la prima volta che esporti un modello, quindi qui basta il pacchetto di base. Questa guida si concentra sulle esportazioni NVIDIA TensorRT, perché TensorRT offre le migliori prestazioni sui dispositivi Jetson.

  1. Aggiorna l'elenco dei pacchetti, installa pip e aggiorna pip all'ultima versione

    sudo apt update
    sudo apt install python3-pip -y
    pip install -U pip
  2. Installa il pacchetto pip ultralytics

    pip install ultralytics
  3. Riavvia il dispositivo

    sudo reboot

Installa PyTorch e Torchvision#

L'installazione di ultralytics descritta sopra installerà Torch e Torchvision. Tuttavia, questi due pacchetti installati tramite pip non sono compatibili con la piattaforma Jetson, basata sull'architettura ARM64, quindi devi installare manualmente un wheel PyTorch precompilato per pip e compilare o installare Torchvision dai sorgenti.

  1. Disinstalla PyTorch e Torchvision attualmente installati

    pip uninstall torch torchvision
  2. Installa torch 2.1.0 e torchvision 0.16.2 in base a JP5.1.2

    pip install https://github.com/ultralytics/assets/releases/download/v0.0.0/torch-2.1.0a0+41361538.nv23.06-cp38-cp38-linux_aarch64.whl
    pip install https://github.com/ultralytics/assets/releases/download/v0.0.0/torchvision-0.16.2+c6f3977-cp38-cp38-linux_aarch64.whl
Nota

Visita la pagina PyTorch per Jetson per accedere a tutte le diverse versioni di PyTorch per le diverse versioni di JetPack. Per un elenco più dettagliato della compatibilità tra PyTorch e Torchvision, visita la pagina sulla compatibilità di PyTorch e Torchvision.

Installa onnxruntime-gpu#

Usa un wheel GPU di ONNX Runtime compatibile con le versioni di JetPack, Python e CUDA. Le versioni attuali di PyPI includono wheel ARM64, ma non sostituiscono i pacchetti specifici del dispositivo per ogni versione di JetPack.

Puoi trovare tutti i pacchetti onnxruntime-gpu disponibili, organizzati per versione di JetPack, versione di Python e altri dettagli di compatibilità, nella matrice di compatibilità di Jetson Zoo ONNX Runtime. Scarica e installa onnxruntime-gpu 1.17.0 con il supporto Python3.8.

wget https://nvidia.box.com/shared/static/zostg6agm00fb6t5uisw51qi6kpcuwzd.whl -O onnxruntime_gpu-1.17.0-cp38-cp38-linux_aarch64.whl
pip install onnxruntime_gpu-1.17.0-cp38-cp38-linux_aarch64.whl
Nota

onnxruntime-gpu aggiorna automaticamente NumPy all'ultima versione; reinstalla quindi NumPy 1.23.5 per risolvere il problema risultante:

pip install numpy==1.23.5

Usa TensorRT su NVIDIA Jetson#

Tra tutti i formati di esportazione dei modelli supportati da Ultralytics, TensorRT offre le prestazioni di inferenza più elevate sui dispositivi NVIDIA Jetson, perciò è la nostra scelta consigliata per le distribuzioni su Jetson. Prima di esportare, installa i binding Python di TensorRT forniti per la tua versione di JetPack e verificali con python3 -c "import tensorrt; print(tensorrt.__version__)". Mantieni il runtime TensorRT 10.x supportato su JetPack 6/7; non sostituirlo con TensorRT 11. Per istruzioni di configurazione e uso avanzato, consulta la nostra guida dedicata all'integrazione di TensorRT.

Puoi anche esportare dal browser senza configurare localmente l'ambiente di compilazione. Nella scheda Esportazione del modello della Ultralytics Platform, seleziona TensorRT e il Jetson di destinazione. Le selezioni Thor vengono convalidate su hardware Thor fisico. Le sei selezioni Orin producono attualmente motori candidati compilati per AGX-Orin; convalidali sul modello Orin previsto prima della distribuzione.

I motori TensorRT sono specifici del dispositivo

TensorRT crea e ottimizza un engine sulla GPU usata per la compilazione. Usa la stessa architettura GPU e lo stesso runtime TensorRT/CUDA del dispositivo di destinazione, quindi convalida ogni engine scaricato sul dispositivo di distribuzione. Le varianti Orin con la stessa architettura non garantiscono automaticamente la portabilità, e la calibrazione INT8 dovrebbe essere eseguita sul dispositivo di destinazione per ottenere i migliori risultati.

Converti il modello in TensorRT ed esegui l'inferenza#

Il modello YOLO26n in formato PyTorch viene convertito in TensorRT per eseguire l'inferenza con il modello esportato.

Esempio
from ultralytics import YOLO

# Carica un modello YOLO26n PyTorch
model = YOLO("yolo26n.pt")

# Esporta il modello in TensorRT
model.export(format="engine")  # crea 'yolo26n.engine'

# Carica il modello TensorRT esportato
trt_model = YOLO("yolo26n.engine")

# Esegui l'inferenza
results = trt_model("https://ultralytics.com/images/bus.jpg")

Usa NVIDIA Deep Learning Accelerator (DLA)#

NVIDIA Deep Learning Accelerator (DLA) è un componente hardware specializzato integrato nei dispositivi NVIDIA Jetson, che ottimizza l'inferenza di deep learning per l'efficienza energetica e le prestazioni. Scaricando le attività dalla GPU (e lasciandola libera per i processi più intensivi), DLA consente ai modelli di funzionare con un consumo energetico inferiore mantenendo un throughput elevato, una caratteristica ideale per i sistemi embedded e le applicazioni di IA in tempo reale.

Compatibilità tra TensorRT e DLA

NVIDIA indica TensorRT 10.7 come l'ultima versione con supporto DLA; TensorRT 11.x non supporta DLA. Usa una versione di TensorRT con supporto DLA compatibile con la tua versione di JetPack. TensorRT 11 non supporta JetPack, nemmeno per le esportazioni che usano solo la GPU.

I seguenti dispositivi Jetson sono dotati di hardware DLA:

Dispositivo JetsonCore DLAFrequenza massima DLA
Serie Jetson AGX Orin21.6 GHz
Jetson Orin NX 16GB2614 MHz
Jetson Orin NX 8GB1614 MHz
Serie Jetson AGX Xavier21.4 GHz
Serie Jetson Xavier NX21.1 GHz
Esempio
from ultralytics import YOLO

# Carica un modello YOLO26n PyTorch
model = YOLO("yolo26n.pt")

# Esporta il modello in TensorRT con DLA abilitato (funziona solo con FP16 o INT8)
model.export(format="engine", device="dla:0", quantize=16)  # dla:0 o dla:1 corrisponde ai core DLA

# Carica il modello TensorRT esportato
trt_model = YOLO("yolo26n.engine")

# Esegui l'inferenza
results = trt_model("https://ultralytics.com/images/bus.jpg")
Nota

Quando usi le esportazioni DLA, alcuni layer potrebbero non essere supportati da DLA e verranno eseguiti dalla GPU. Questo fallback può introdurre latenza aggiuntiva e influire sulle prestazioni complessive dell'inferenza. Pertanto, rispetto a TensorRT eseguito interamente sulla GPU, DLA non è progettato principalmente per ridurre la latenza dell'inferenza. Il suo scopo principale è invece aumentare il throughput e migliorare l'efficienza energetica.

Benchmark NVIDIA Jetson YOLO26#

Il team Ultralytics ha eseguito benchmark di YOLO26 su 11 formati di modello diversi, misurando velocità e precisione: PyTorch, TorchScript, ONNX, OpenVINO, TensorRT, TF SavedModel, TF GraphDef, TF Lite, MNN, NCNN, ExecuTorch. I benchmark sono stati eseguiti su NVIDIA Jetson AGX Thor Developer Kit, NVIDIA Jetson AGX Orin Developer Kit (64GB), NVIDIA Jetson Orin Nano Super Developer Kit e Seeed Studio reComputer J4012 con Jetson Orin NX 16GB, a precisione FP32 e con la dimensione predefinita dell'immagine di input pari a 640.

Grafici comparativi#

Anche se tutte le esportazioni dei modelli funzionano su NVIDIA Jetson, il grafico comparativo qui sotto include solo PyTorch, TorchScript e TensorRT. Puoi trovare i benchmark di tutte le esportazioni nella sezione successiva al grafico.

Kit per sviluppatori NVIDIA Jetson AGX Thor#

Jetson AGX Thor Benchmarks
Testato con Ultralytics 8.3.226

Kit per sviluppatori NVIDIA Jetson AGX Orin (64GB)#

Jetson AGX Orin Benchmarks
Testato con Ultralytics 8.4.32

Kit per sviluppatori NVIDIA Jetson Orin Nano Super#

Jetson Orin Nano Super Benchmarks
Testato con Ultralytics 8.4.33

NVIDIA Jetson Orin NX 16GB#

Jetson Orin NX 16GB Benchmarks
Testato con Ultralytics 8.4.33

Tabelle di confronto dettagliate#

La tabella seguente mostra i risultati dei benchmark per cinque diversi modelli YOLO26 (YOLO26n, YOLO26s, YOLO26m, YOLO26l, YOLO26x) in 11 formati diversi (PyTorch, TorchScript, ONNX, OpenVINO, TensorRT, TF SavedModel, TF GraphDef, TF Lite, MNN, NCNN, ExecuTorch), indicando lo stato, le dimensioni, la metrica mAP50-95(B) e il tempo di inferenza per ogni combinazione.

Kit per sviluppatori NVIDIA Jetson AGX Thor#

Prestazioni
FormatoStatoDimensioni su disco (MB)mAP50-95(B)Tempo di inferenza (ms/im)
PyTorch✅5.30.47987.39
TorchScript✅9.80.47894.21
ONNX✅9.50.47676.58
OpenVINO✅10.10.479417.50
TensorRT (FP32)✅13.90.47911.90
TensorRT (FP16)✅7.60.47971.39
TensorRT (INT8)✅6.50.42731.52
TF SavedModel✅25.70.476447.24
TF GraphDef✅9.50.476445.98
TF Lite✅9.90.4764182.04
MNN✅9.40.478421.83

Benchmark eseguito con Ultralytics 8.4.7

Nota

Il tempo di inferenza non include la pre/post-elaborazione.

Kit per sviluppatori NVIDIA Jetson AGX Orin (64GB)#

Prestazioni
FormatoStatoDimensioni su disco (MB)mAP50-95(B)Tempo di inferenza (ms/im)
PyTorch✅5.30.479011.58
TorchScript✅9.80.47704.60
ONNX✅9.50.47709.87
OpenVINO✅9.60.482028.80
TensorRT (FP32)✅11.50.47704.18
TensorRT (FP16)✅7.90.47892.62
TensorRT (INT8)✅5.40.46402.30
TF SavedModel✅24.60.476071.10
TF GraphDef✅9.50.476070.02
TF Lite✅9.90.4760227.94
MNN✅9.40.476032.46
NCNN✅9.30.481029.93

Benchmark eseguito con Ultralytics 8.4.32

Nota

Il tempo di inferenza non include la pre/post-elaborazione.

Kit per sviluppatori NVIDIA Jetson Orin Nano Super#

Prestazioni
FormatoStatoDimensioni su disco (MB)mAP50-95(B)Tempo di inferenza (ms/im)
PyTorch✅5.30.479015.60
TorchScript✅9.80.477012.60
ONNX✅9.50.476015.76
OpenVINO✅9.60.482056.23
TensorRT (FP32)✅11.30.47707.53
TensorRT (FP16)✅8.10.48004.57
TensorRT (INT8)✅5.30.44903.80
TF SavedModel✅24.60.4760118.33
TF GraphDef✅9.50.4760116.30
TF Lite✅9.90.4760286.00
MNN✅9.40.476068.77
NCNN✅9.30.481047.50

Benchmark eseguito con Ultralytics 8.4.33

Nota

Il tempo di inferenza non include la pre/post-elaborazione.

NVIDIA Jetson Orin NX 16GB#

Prestazioni
FormatoStatoDimensioni su disco (MB)mAP50-95(B)Tempo di inferenza (ms/im)
PyTorch✅5.30.479913.90
TorchScript✅9.80.478711.60
ONNX✅9.50.476314.18
OpenVINO✅9.60.481940.19
TensorRT (FP32)✅11.40.47707.01
TensorRT (FP16)✅8.00.47894.13
TensorRT (INT8)✅5.50.44893.49
TF SavedModel✅24.60.476492.34
TF GraphDef✅9.50.476492.06
TF Lite✅9.90.4764254.43
MNN✅9.40.476048.55
NCNN✅9.30.480534.31

Benchmark eseguito con Ultralytics 8.4.33

Nota

Il tempo di inferenza non include la pre/post-elaborazione.

Scopri altri benchmark condotti da Seeed Studio su diverse versioni dell'hardware NVIDIA Jetson.

Riproduci i nostri risultati#

Per riprodurre i benchmark Ultralytics qui sopra con tutti i formati di esportazione, esegui questo codice:

Esempio
from ultralytics import YOLO

# Carica un modello YOLO26n PyTorch
model = YOLO("yolo26n.pt")

# Esegui il benchmark di velocità e accuratezza di YOLO26n sul dataset COCO128 per tutti i formati di esportazione
results = model.benchmark(data="coco128.yaml", imgsz=640)

I risultati dei benchmark possono variare in base alla configurazione hardware e software esatta del sistema e al carico di lavoro del sistema al momento dell'esecuzione dei benchmark. Per ottenere risultati più affidabili, usa un dataset con un numero elevato di immagini, ad es. data='coco.yaml' (5000 immagini di convalida).

Best practice per l'utilizzo di NVIDIA Jetson#

Segui queste best practice per ottenere le massime prestazioni da YOLO26 su NVIDIA Jetson.

  1. Attiva la modalità di alimentazione MAX

    L'attivazione della modalità MAX Power su Jetson garantisce che tutti i core CPU e GPU siano attivi.

    sudo nvpmodel -m 0
  2. Attiva Jetson Clocks

    L'attivazione di Jetson Clocks garantisce che tutti i core CPU e GPU funzionino alla frequenza massima.

    sudo jetson_clocks
  3. Installa l'applicazione Jetson Stats

    Usa l'applicazione jetson-stats per monitorare le temperature dei componenti del sistema, visualizzare l'utilizzo di CPU, GPU e RAM, cambiare le modalità di alimentazione, impostare le frequenze massime e controllare le informazioni su JetPack.

    sudo apt update
    sudo pip install jetson-stats
    sudo reboot
    jtop
jtop dashboard from jetson-stats showing Jetson CPU, GPU, and memory utilization

Suggerimenti per ottimizzare la memoria su NVIDIA Jetson#

La memoria disponibile è spesso il fattore limitante sui dispositivi Jetson, in particolare sulle varianti con meno memoria, come Jetson Orin Nano (8 GB) o Orin NX 8 GB. I suggerimenti seguenti sono modifiche pratiche e a basso rischio che, nel loro insieme, possono liberare diverse centinaia di megabyte e consentirti di eseguire modelli YOLO più grandi o supportare ulteriori carichi di lavoro paralleli. Per una trattazione completa, consulta il blog NVIDIA sull'ottimizzazione dell'efficienza della memoria su Jetson.

1. Passa all'avvio headless (senza interfaccia grafica)#

Se il tuo Jetson è collegato tramite SSH o viene usato come dispositivo di produzione senza display, eliminare l'ambiente desktop e il server grafico può liberare fino a 865 MB di RAM:

sudo systemctl set-default multi-user.target
sudo reboot

Per ripristinare in seguito l'ambiente desktop:

sudo systemctl set-default graphical.target
sudo reboot

2. Disattiva i servizi di sistema inutilizzati#

I servizi in background non essenziali (Bluetooth, gestori della connettività, demoni hardware inutilizzati) consumano complessivamente circa 32 MB. Elenca i servizi attivi e disattiva quelli non necessari per la tua distribuzione:

# List running services
systemctl list-units --type=service --state=running

# Disable a service
sudo systemctl disable SERVICE_NAME

3. Analizza l'utilizzo della memoria#

Prima di ottimizzare, individua i processi che consumano effettivamente RAM. procrank ordina i processi in base a PSS (dimensione proporzionale del set), che riflette l'impronta di memoria effettiva per processo con maggiore precisione rispetto a RSS (dimensione del set residente, ovvero il totale delle pagine di RAM fisica mappate da un processo, incluse quelle condivise con altri processi):

git clone https://github.com/csimmonds/procrank_linux.git
cd procrank_linux && make
sudo ./procrank

Per visualizzare le allocazioni GPU e NvMap (pipeline CUDA/video) per processo:

sudo cat /sys/kernel/debug/nvmap/iovmm/clients

4. Esegui l'inferenza senza display in produzione#

Per le pipeline di inferenza che non richiedono un'anteprima in tempo reale, disattivare i componenti relativi al display (Tiler, OSD, DisplaySink) può far risparmiare oltre 200 MB solo dalla pipeline. Con Ultralytics YOLO, disattiva il visualizzatore e salva invece i risultati su disco:

Esempio
from ultralytics import YOLO

model = YOLO("yolo26n.engine")

# show=False impedisce l'apertura di qualsiasi finestra di visualizzazione; save=True scrive su disco i risultati annotati
results = model.predict(source="video.mp4", show=False, save=True)

Impatto complessivo#

OttimizzazioneMemoria risparmiata (circa)
Disattiva l'interfaccia grafica desktop~865 MB
Disattiva i servizi del sistema operativo inutilizzati~32 MB
Pipeline di inferenza headless (senza display)~200+ MB
Totale (interventi semplici)~1 GB+

Combinare queste modifiche è particolarmente utile quando si usano modelli TensorRT INT8 su dispositivi con memoria limitata: può fare la differenza tra riuscire o meno a caricare in memoria una variante di modello più grande.

Prossimi passaggi#

Per approfondire e ricevere supporto, consulta la documentazione Ultralytics YOLO26.

Domande frequenti#

  • Per distribuire Ultralytics YOLO26 sui dispositivi NVIDIA Jetson, esegui prima il flashing del dispositivo Jetson con NVIDIA JetPack SDK. Poi, usa un'immagine Docker precompilata per una configurazione rapida oppure installa manualmente i pacchetti richiesti.

  • I modelli YOLO26 sono stati sottoposti a benchmark sui dispositivi NVIDIA Jetson AGX Thor, AGX Orin, Orin Nano Super e Orin NX; il formato TensorRT offre le migliori prestazioni di inferenza. La sezione Tabelle comparative dettagliate elenca mAP50-95 e il tempo di inferenza per ogni modello e formato di esportazione.

  • TensorRT è consigliato per distribuire i modelli YOLO26 su NVIDIA Jetson perché offre l'inferenza più rapida tra tutti i formati di esportazione, ottimizzando il modello per la GPU Jetson.

  • Per installare PyTorch e Torchvision su NVIDIA Jetson, disinstalla prima le eventuali versioni esistenti installate tramite pip. Poi installa manualmente le versioni compatibili di PyTorch e Torchvision per l'architettura ARM64 di Jetson. Trovi le istruzioni dettagliate per questa procedura nella sezione Installa PyTorch e Torchvision.

  • Per massimizzare le prestazioni di YOLO26 su NVIDIA Jetson, segui queste best practice:

    1. Attiva la modalità di alimentazione MAX per usare tutti i core CPU e GPU.
    2. Attiva Jetson Clocks per eseguire tutti i core alla frequenza massima.
    3. Installa l'applicazione Jetson Stats per monitorare le metriche del sistema.

    Per i comandi e ulteriori dettagli, consulta la sezione Best practice per l'utilizzo di NVIDIA Jetson.

  • La RAM disponibile è spesso il collo di bottiglia sui dispositivi Jetson con meno memoria. Ecco tre semplici interventi che, insieme, possono liberare oltre 1 GB:

    1. Passa all'avvio headless (sudo systemctl set-default multi-user.target) per eliminare l'interfaccia grafica desktop (~865 MB liberati).
    2. Disattiva i servizi inutilizzati, come Bluetooth o i gestori della connettività (~32 MB liberati).
    3. Esegui l'inferenza senza display impostando show=False nella chiamata YOLO predict, evitando così di allocare memoria per la pipeline del display (~200+ MB liberati).

    Usa procrank per analizzare l'utilizzo della RAM per processo e sudo cat /sys/kernel/debug/nvmap/iovmm/clients per esaminare le allocazioni GPU. Consulta la sezione Suggerimenti per ottimizzare la memoria per tutti i dettagli.

  • TensorRT 10.3.0 incluso in JetPack 6 presenta un problema noto che impedisce di compilare motori INT8 con output senza NMS (nms=False) o con NMS integrata (nms=True). Quando Ultralytics rileva questa combinazione, seleziona la testa one-to-many per le esportazioni senza NMS e interrompe le esportazioni nms=True con un errore.

    Per ripristinare queste esportazioni INT8, aggiorna TensorRT a una versione più recente (ad es. 10.7.0 o successiva):

    wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/arm64/cuda-keyring_1.1-1_all.deb
    sudo dpkg -i cuda-keyring_1.1-1_all.deb
    sudo apt-get update
    sudo apt-get install -y tensorrt

    Dopo l'aggiornamento, esegui di nuovo l'esportazione. Per maggiori dettagli, consulta il problema GitHub #23841 e il problema #26655.

Commenti194

Fai una domanda, condividi i tuoi risultati o aiuta qualcun altro.

  1. pknowlerGitHub

    Struggling with this on a new Jetson Orin Nano JP6.1. All good following the steps and installs to give torch.cuda.is_available as TRUE. The TensorRT export fails though with a number of dependency errors and I can’t re-install it with anything. Any pointers please?

    1
    1. Glenn JocherUltralytics

      @pknowler it seems like the dependency errors during TensorRT export might stem from compatibility issues. Please ensure you’ve installed the required JetPack-compatible versions of PyTorch, Torchvision, and TensorRT as outlined in the Jetson Quick Start Guide. Additionally, verify that onnxruntime-gpu and cuSPARSELt are correctly installed per the guide. If the issue persists, you can try reinstalling TensorRT using the NVIDIA SDK Manager for your JetPack version. Let me know if you need further clarification!

    2. pknowlerGitHub

      Thank you. I will try again working through all of the steps and let you know how it goes!

    3. pknowlerGitHub

      I’ve tried again but this as far as I get… I have followed everything carefully.

      WARNING ⚠️ TensorRT requires GPU export, automatically assigning device=0 Ultralytics 8.3.57 🚀 Python-3.10.16 torch-2.5.0a0+872d972e41.nv24.08 CUDA:0 (Orin, 7620MiB) YOLO11n summary (fused): 238 layers, 2,616,248 parameters, 0 gradients, 6.5 GFLOPs

      PyTorch: starting from 'yolo11n.pt' with input shape (1, 3, 640, 640) BCHW and output shape(s) (1, 84, 8400) (5.4 MB)

      ONNX: starting export with onnx 1.17.0 opset 19... ONNX: slimming with onnxslim 0.1.46... ONNX: export success ✅ 3.1s, saved as 'yolo11n.onnx' (10.2 MB) requirements: Ultralytics requirement ['tensorrt>7.0.0,!=10.1.0'] not found, attempting AutoUpdate... error: subprocess-exited-with-error

      × python setup.py egg_info did not run successfully. │ exit code: 1 ╰─> [6 lines of output] Traceback (most recent call last): File "", line 2, in File "", line 34, in File "/tmp/pip-install-d0ljzl60/tensorrt-cu12_db0552ac431148489703f6f75df6764f/setup.py", line 71, in raise RuntimeError("TensorRT does not currently build wheels for Tegra systems") RuntimeError: TensorRT does not currently build wheels for Tegra systems [end of output]

      note: This error originates from a subprocess, and is likely not a problem with pip. error: metadata-generation-failed

      × Encountered error while generating package metadata. ╰─> See above for output.

      note: This is an issue with the package mentioned above, not pip. hint: See above for details. Retry 1/2 failed: Command 'pip install --no-cache-dir "tensorrt>7.0.0,!=10.1.0" ' returned non-zero exit status 1. error: subprocess-exited-with-error

      × python setup.py egg_info did not run successfully. │ exit code: 1 ╰─> [6 lines of output] Traceback (most recent call last): File "", line 2, in File "", line 34, in File "/tmp/pip-install-782zjqf9/tensorrt-cu12_44d261aea7c5489db3e9ce15cec8ee64/setup.py", line 71, in raise RuntimeError("TensorRT does not currently build wheels for Tegra systems") RuntimeError: TensorRT does not currently build wheels for Tegra systems [end of output]

      note: This error originates from a subprocess, and is likely not a problem with pip. error: metadata-generation-failed

      × Encountered error while generating package metadata. ╰─> See above for output.

      note: This is an issue with the package mentioned above, not pip. hint: See above for details.

    4. Glenn JocherUltralytics

      @pknowler it seems your TensorRT version is not compatible with the Jetson's ARM64 architecture. TensorRT does not provide pre-built Python wheels for Tegra systems. Please ensure you have TensorRT installed via the NVIDIA SDK Manager as part of the JetPack installation, and confirm compatibility with your JetPack version. The Jetson Quick Start Guide might also help clarify setup specifics. Let me know if further assistance is needed!

    5. pknowlerGitHub

      @glenn-jocher thank you. I have found the issue now. I am using miniconda and for some reason TensorRT Python bindings are not available in conda envs. So I can import Torch with no problem but the import TensorRT fails. I have re-installed without conda and all works as expected! 👍🏻

    1. Paula DerrengerUltralytics(modificato)

      In terms of reducing inference time on your NVIDIA Jetson device, using TensorRT is generally recommended. TensorRT is specifically designed to provide high performance for NVIDIA hardware by optimizing deep learning models for inference. It accelerates models by fusing layers, pruning unused layers, and quantizing variables, among other optimizations. 🚀

      DeepStream, on the other hand, is more focused on streaming analytics and can utilize TensorRT-optimized models. For most cases, directly using TensorRT for model optimization is your go-to for minimizing inference times.

      Here's a quick example of exporting a YOLO model to TensorRT and running inference, tailored for your convenience:

      from ultralytics import YOLO
      
      # Load your YOLO model
      model = YOLO('yolo26n.pt')
      
      # Export to TensorRT format
      model.export(format='engine')
      
      # Load the TensorRT model
      trt_model = YOLO('yolo26n.engine')
      
      # Run inference using the TensorRT model
      results = trt_model('your_image_or_video.jpg')

      Feel free to explore more about deploying on NVIDIA Jetson with our Quick Start Guide for NVIDIA Jetson. Happy inferencing! 🌟

      2
    2. Paula DerrengerUltralytics(modificato)

      Hey there! 😊 It looks like you're interested in speeding up your inference times. Between TensorRT and DeepStream on NVIDIA Jetson devices, TensorRT typically offers more direct optimization for reducing inference time for individual models by providing lower-level optimizations and quantization options.

      DeepStream is a comprehensive framework for building end-to-end video analytics applications, which utilizes TensorRT under the hood for model inference. DeepStream excels in scenarios involving video streams, multi-stream input, and complex pipelines involving multiple AI models and analytics. If your primary goal is solely to minimize inference time for a specific model, direct optimization via TensorRT can provide more focused improvements.

      A quick example of exporting a YOLO model to TensorRT:

      from ultralytics import YOLO
      
      # Load your YOLO model
      model = YOLO('yolo26n.pt')
      
      # Export to TensorRT format
      model.export(format='engine')  # This creates 'yolo26n.engine'
      
      # Done! Now you can use 'yolo26n.engine' for faster inference.

      However, if your application benefits from a broader video analytics pipeline, incorporating DeepStream into your workflow can optimize and streamline these processes on top of efficiently handling inference with TensorRT optimized models. 🚀

      Hope this gives you a clearer picture! If you have more questions, feel free to ask.

      1
  2. Yufeng Wu(modificato)

    Hi, ultralytics team. I'm trying yolo26n-pose on Jetson orin agx 64gb and have this problem: I tried building engine with dynamic=true but failed, with the log saying ForeignNode[/model.23/Sigmoid.../model.23/Flatten] (Myelin) Skipping tactic ... Incompatible effective shapes in operation /model_23/Sub between: Unsqueeze_18_output_0 f32[1,2,33600] and Slice_output_0 f32[1,2,21504] Could not find any implementation for node {ForeignNode[/model.23/Sigmoid.../model.23/Flatten]} Is there a way to fix it?

    1. Glenn JocherUltralytics

      This indicates a TensorRT dynamic-shape profile failure in the pose head. As a workaround, export a fixed-resolution engine, e.g. yolo export model=yolo26n-pose.pt format=engine imgsz=640 batch=1 dynamic=False; build separate engines if multiple resolutions are required. If dynamic shapes are essential, please open a new issue with the full text log, yolo checks output, and JetPack/TensorRT/Ultralytics versions so we can reproduce it; see the TensorRT export guide.

  3. Haotian Gong(modificato)

    Hi, everyone,

    I'd like to flag two issues with the Jetson installation guide that may need updating.


    1. PyTorch 2.10.0 / Torchvision 0.25.0 doesn't work on JetPack 6.1

    Following the "Start with naive installation" and "Run on JetPack 6.1" sections, I installed Pytorch 2.10.0 and Torchvision 0.25.0 as instructed.

    Running my script then throws:

    ImportError: libcublas.so.13: cannot open shared object file: No such file or directory

    So I have to install Pytorch 2.8.0 with the guide:

    https://forums.developer.nvidia.com/t/jetson-orin-nano-issues-running-yolo-ultralytics-due-to-pytorch-torchvision-compatibility-jetpack-6-x-cuda-12-6/365683/9

    And this version of Pytorch works. So maybe is the guide on Jetson obsolete or mistaken?


    2. CSI camera requires the JetPack-bundled OpenCV, not the PyPI one

    On Jetson, using a CSI camera (via nvarguscamerasrc) requires the OpenCV that ships with JetPack, because the PyPI opencv-python wheel on aarch64 is built without GStreamer support.

    However, pip install ultralytics pulls in opencv-python as a dependency, which shadows the JetPack-bundled version and silently breaks CSI camera pipelines.

    pip uninstall -y opencv-python opencv-python-headless
    sudo apt install -y python3-opencv   # JetPack-bundled OpenCV with GStreamer

    Thanks!

    1. Lakshantha DissanayakeUltralytics

      Hi @ZephyrKeXiner Could you please double check whether you are using JetPack 6.1 system with CUDA 12.6? I have just verified on my JetPack 6.1 system with CUDA 12.6 with torch-2.10.0 and torchvision-0.25.0 as mentioned in the Ultralytics Jetson Guide and it works well.

      For JetPack 6.1, output of cat /etc/nv_tegra_release should look like below:

      (env) nvidia@ubuntu:~/yolo26/ultralytics$ cat /etc/nv_tegra_release 
      # R36 (release), REVISION: 4.3, GCID: 38968081, BOARD: generic, EABI: aarch64, DATE: Wed Jan  8 01:49:37 UTC 2025
      # KERNEL_VARIANT: oot
      TARGET_USERSPACE_LIB_DIR=nvidia
      TARGET_USERSPACE_LIB_DIR_PATH=usr/lib/aarch64-linux-gnu/nvidia

      Could you also share yours please? Thanks!

      cc: @raimbekovm

      1
    2. Haotian Gong

      Hi @lakshanthad,

      My output of cat /etc/nv_tegra_release is below:

      $ cat /etc/nv_tegra_release 
      # R36 (release), REVISION: 5.0, GCID: 43688277, BOARD: generic, EABI: aarch64, DATE: Fri Jan 16 03:50:45 UTC 2026
      # KERNEL_VARIANT: oot
      TARGET_USERSPACE_LIB_DIR=nvidia
      TARGET_USERSPACE_LIB_DIR_PATH=usr/lib/aarch64-linux-gnu/nvidia

      I notice that the revision is 5.0 not 4.3. Maybe does it cause the error?

  4. smbunnGitHub(modificato)

    I would have thought the Jetson Orin Nano could do more than 5 streams, the deepstream examples with cars handles 16 streams of 1080p at what appears to be full framerate.

    As a control systems engineer by training, edge devices seem a much better way to go so I prefer your jetson design. Even if one Jetson has an issue you are still processing 45 streams whereas a problem on your centralised server in option 2 takes everything down.

    Are you going to set up an MQTT feed for alerts from each stream? Maybe consider secondary processing and a buffer, so the edge device detects what it thinks is an issue and a centralised server grabs 5 seconds of video of the incident and double checks.

    Simon BUNN

    General


    From: Yash Patel @.> Sent: Friday, 22 May 2026 4:00 am To: ultralytics/ultralytics @.> Cc: Bunn, Simon @.>; Mention @.> Subject: Re: [ultralytics/ultralytics] guides/nvidia-jetson/ (Discussion #9942)

    Hi, I am designing an AI-based CCTV surveillance system for a residential society with 50 RTSP camera streams. My current architecture uses multiple NVIDIA Jetson Orin Nano Super Developer Kit devices for distributed inference processing. I ZjQcmQRYFpfptBannerStart Attention The sender's identity could not be verified and someone may be impersonating the sender. L'identité de l'expéditeur n'a pas pu être vérifiée et quelqu'un pourrait se faire passer pour l'expéditeur.

    ZjQcmQRYFpfptBannerEnd

    Hi, I am designing an AI-based CCTV surveillance system for a residential society with 50 RTSP camera streams. My current architecture uses multiple NVIDIA Jetson Orin Nano Super Developer Kit devices for distributed inference processing. I estimate that one Jetson Orin Nano Super (8GB) can process around 5 RTSP streams at usable real-time FPS. This requires approximately 10 Jetson devices for the full deployment. Each device costs around ₹33,000 (~USD 395). The total estimated hardware cost is ₹330,000 (~USD 3,950). I want to compare this setup against a centralized Linux-based custom PC/server with NVIDIA GPU acceleration under the same budget. The comparison focuses on AI inference performance, scalability, reliability, and 24/7 operational efficiency. The workload mainly includes multi-stream YOLO/object detection inference on live RTSP feeds. I want to determine which architecture provides the best computation-to-cost ratio for this surveillance use case.

    — Reply to this email directly, view it on GitHubhttps://urldefense.com/v3/__https://github.com/orgs/ultralytics/discussions/9942*discussioncomment-17006943__;Iw!!ElGdukoduuk!URiPjKkgDoMT_k2Y6pEd3XMEBGMDjhwgtQmwWXR9F4R4L7ymNloZD79NanOjYVg9YYmaARqyZGbgFv7U15KzgdcU$, or unsubscribehttps://urldefense.com/v3/__https://github.com/notifications/unsubscribe-auth/ABAAMILYUP7PPODNHDR3HC3434RYLAVCNFSM6AAAAABGAK2IGGVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTOMBQGY4TIMY__;!!ElGdukoduuk!URiPjKkgDoMT_k2Y6pEd3XMEBGMDjhwgtQmwWXR9F4R4L7ymNloZD79NanOjYVg9YYmaARqyZGbgFv7U1zjhi4YU$. Triage notifications on the go with GitHub Mobile for iOShttps://urldefense.com/v3/__https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675__;!!ElGdukoduuk!URiPjKkgDoMT_k2Y6pEd3XMEBGMDjhwgtQmwWXR9F4R4L7ymNloZD79NanOjYVg9YYmaARqyZGbgFv7U13gOq1qb$ or Androidhttps://urldefense.com/v3/__https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign*3Dnotification-email*26utm_medium*3Demail*26utm_source*3Dgithub__;JSUlJSU!!ElGdukoduuk!URiPjKkgDoMT_k2Y6pEd3XMEBGMDjhwgtQmwWXR9F4R4L7ymNloZD79NanOjYVg9YYmaARqyZGbgFv7U1zOl3XaF$. You are receiving this because you were mentioned.Message ID: @.***>

    1. Glenn JocherUltralytics

      5 streams on an Orin Nano Super is just a conservative YOLO/Python estimate; with a headless TensorRT or DeepStream pipeline, MAX power mode, and jetson_clocks, higher 1080p stream counts are definitely possible and depend heavily on the model size, input resolution, decode/encode, and any tracking or alert logic you add, so please drop your exact model, imgsz, FPS target, and RTSP resolution here and we can help size it more accurately using the Jetson deployment guide and DeepStream Jetson guide.

  5. yashpatel2010GitHub

    Hi, I am designing an AI-based CCTV surveillance system for a residential society with 50 RTSP camera streams. My current architecture uses multiple NVIDIA Jetson Orin Nano Super Developer Kit devices for distributed inference processing. I estimate that one Jetson Orin Nano Super (8GB) can process around 5 RTSP streams at usable real-time FPS. This requires approximately 10 Jetson devices for the full deployment. Each device costs around ₹33,000 (~USD 395). The total estimated hardware cost is ₹330,000 (~USD 3,950). I want to compare this setup against a centralized Linux-based custom PC/server with NVIDIA GPU acceleration under the same budget. The comparison focuses on AI inference performance, scalability, reliability, and 24/7 operational efficiency. The workload mainly includes multi-stream YOLO/object detection inference on live RTSP feeds. I want to determine which architecture provides the best computation-to-cost ratio for this surveillance use case.

  6. RotatingGiraffeGitHub

    Hi there, I am trying to run a fine-tuned YOLO26 based model for an OBB task on a Xavier NX running JetPack 5. I am using your container (ultralytics/ultralytics:latest-jetson-jetpack5) and trying to compile from PyTorch to TensorRT. I get the error: "mymodel.pt requires numpy>=1.26.1, however numpy==1.23.5 is installed."

    Previously I have successfully the file on my host machine from PT to ONNX, copied ONNX to Jetson and then used tensorrtexec to compile to Engine. But the resulting models did not create any bounding boxes.

    I wanted to ask if JetPack 5 even supports Yolo26 or if the model contains some incompatible layers? YOLO11 OBB works fine which I find extra-weird. Or is there likely an error somewhere else in my script?

    1. Glenn JocherUltralytics

      Thanks for posting this here—JetPack 5 on Xavier NX is supported for YOLO26/TensorRT in our Jetson guide, so the numpy>=1.26.1 error points more to an environment/checkpoint mismatch than unsupported OBB layers; please paste the full text traceback and yolo checks output, and note that an engine built with tensorrtexec can still produce empty OBB results if the Ultralytics export metadata/postprocessing is not preserved.

    2. RotatingGiraffeGitHub

      Hi there, I tried to create a minimal reproducible example for you, based on what I am doing. I hope it helps with debugging.

      The yolo checks you asked for is in the script output, along with the export script. I've also included the bit where I export via trtexex just for the sake of completion - it might very well be possible that I omit required metadata with this. My prefered goal would be that everything works via the Ultralytics Python API, but I had to search for workaround due to the NumPy error when compiling PT files.

      I am now using the Ultralytics base image in the MRE which I hoped would be easier to debug for us. The NumPy error persists on the new image. Previously I was using something built upon l4t-ml.

      Additionally, ONNX fails with an error ( `GLIBCXX_3.4.29' not found I think it was) but that is not a concern to me since I engine would be my prefered way to go. I have not installed any additional packages and kept it very vanilla (see dockerfile). The MRE also weirdly produced a new error during engine inference, but I'd say lets just focus on getting the Ultralytics native pipeline to work and then maybe all the other issues were just from my workarounds and disappear.

      Execution Command
      sudo docker compose up --build
      
      Directory Structure
      obb-test-repro/
       ┣ 📂data/
       ┃ ┗ 📜mydataset.yaml
       ┣ 📂models/
       ┃ ┣ 📜my-yolo26m-model.onnx
       ┃ ┗ 📜my-yolo26m-model.pt
       ┣ 📂src/
       ┃ ┗ 📜main.py
       ┣ 📜docker-compose.yml
       ┣ 📜Dockerfile
       ┗ 📜requirements.txt
      
      Dockerfile
      FROM ultralytics/ultralytics:latest-jetson-jetpack5
      
      WORKDIR /app
      
      # Copy the test app
      COPY src/ ./src/
      
      # Run the test
      CMD ["python3", "src/main.py"]
      
      docker-compose.yml
      version: '3.8'
      
      services:
        obb-test-jp5:
          image: obb-test:latest
          container_name: obb_test_jp5
          build:
            context: .
            dockerfile: Dockerfile
          runtime: nvidia
          network_mode: host
          ipc: host
          restart: "no"
          volumes:
            - ./models:/app/models
            - ./data:/app/data
          environment:
            - NVIDIA_VISIBLE_DEVICES=all
            - NVIDIA_DRIVER_CAPABILITIES=all
      
      main.py
      """
      Minimal OBB model test script.
      
      Behavior:
      - Runs `yolo checks` to output environment details.
      - Uses files in `/app/models` (container path mapped from `obb-test/models`).
      - Expects two files (hard-coded):
        - `/app/models/my-yolo26m-model.pt` (optional) — used for Ultralytics native export
        - `/app/models/my-yolo26m-model.onnx` (optional) — used for trtexec compilation
      - Runs both export methods (if files exist) and validates inference on each compiled engine using mydataset dataset.
      """
      
      import logging
      import sys
      import subprocess
      import os
      from pathlib import Path
      from ultralytics import YOLO
      
      # Configure logging
      logging.basicConfig(
          level=logging.INFO,
          format='%(asctime)s - %(levelname)s - %(message)s'
      )
      logger = logging.getLogger(__name__)
      
      
      def run_yolo_checks():
          """Execute yolo checks via CLI to log environment details."""
          logger.info("=" * 60)
          logger.info("Environment Verification (yolo checks)")
          logger.info("=" * 60)
          try:
              res = subprocess.run(["yolo", "checks"], capture_output=True, text=True, check=True)
              logger.info("\n" + res.stdout.strip())
          except subprocess.CalledProcessError as e:
              logger.error(f"yolo checks failed:\n{e.stderr}")
          except FileNotFoundError:
              logger.error("yolo CLI command not found. Ensure ultralytics is installed in this environment.")
      
      
      def test_export_to_engine():
          models_dir = Path("/app/models")
          pt_path = models_dir / "my-yolo26m-model.pt"
          onnx_path = models_dir / "my-yolo26m-model.onnx"
          dataset_yaml = "/app/data/mydataset.yaml"
      
          logger.info("=" * 60)
          logger.info("Minimal OBB Test - Export & Validation")
          logger.info("=" * 60)
      
          engines_tested = []
      
          # Part 1: Ultralytics native export (PT -> Engine)
          if pt_path.exists():
              try:
                  logger.info(f"\nRunning Ultralytics native export for: {pt_path.name}")
                  model = YOLO(str(pt_path))
                  exported = model.export(format="engine", imgsz=1024, device=0, half=True)
                  
                  # Resolve the exported engine path
                  native_engine = str(exported) if isinstance(exported, (str, Path)) else str(pt_path.with_suffix('.engine'))
                  
                  if os.path.exists(native_engine):
                      logger.info(f"Ultralytics export successfully produced: {native_engine}")
                      engines_tested.append(native_engine)
                  else:
                      logger.error(f"Ultralytics export completed, but engine file was not found at {native_engine}")
              except Exception as e:
                  logger.error(f"Ultralytics native export failed: {e}", exc_info=True)
          else:
              logger.info(f"PT file not found ({pt_path}); skipping Ultralytics native export.")
      
          # Part 2: trtexec compilation (ONNX -> Engine)
          if onnx_path.exists():
              try:
                  logger.info(f"\nRunning trtexec to compile ONNX: {onnx_path.name}")
                  trt_engine = str(onnx_path.with_suffix('.trt.engine'))
                  cmd = [
                      "/usr/src/tensorrt/bin/trtexec",
                      f"--onnx={onnx_path}",
                      f"--saveEngine={trt_engine}",
                      "--fp16",
                      "--memPoolSize=workspace:768", # Prevents Out of Memory
                  ]
                  logger.info(f"Running command: {' '.join(cmd)}")
                  res = subprocess.run(cmd, capture_output=True, text=True)
                  
                  if res.returncode != 0:
                      logger.error("trtexec compilation failed.")
                      logger.error(f"trtexec stderr:\n{res.stderr}")
                  else:
                      if os.path.exists(trt_engine):
                          logger.info(f"trtexec successfully saved engine to {trt_engine}")
                          engines_tested.append(trt_engine)
                      else:
                          logger.error(f"trtexec returned success, but engine file was not found at {trt_engine}")
              except Exception as e:
                  logger.error(f"trtexec execution failed: {e}", exc_info=True)
          else:
              logger.info(f"ONNX file not found ({onnx_path}); skipping trtexec compilation.")
      
          if not engines_tested:
              logger.error("\nNo engines are available to test. Exiting validation phase.")
              return False
      
          # Part 3: Inference Validation
          validation_success = True
          for eng in engines_tested:
              try:
                  logger.info(f"\nValidating inference with engine: {eng}")
                  
                  model = YOLO(eng, task="obb")
                  metrics = model.val(data=dataset_yaml, imgsz=1024, batch=1, verbose=False)
                  
                  logger.info(f"Validation metrics for {eng}: mAP50-95={metrics.box.map:.4f}, mAP50={metrics.box.map50:.4f}")
              except Exception as e:
                  logger.error(f"Validation failed on {eng}: {e}", exc_info=True)
                  validation_success = False
      
          logger.info("\nExecution complete.")
          return validation_success
      
      
      if __name__ == "__main__":
          run_yolo_checks()
          
          try:
              success = test_export_to_engine()
              logger.info("\n" + "=" * 60)
              if success:
                  logger.info("✓ Test passed")
              else:
                  logger.info("✗ Test failed")
              logger.info("=" * 60)
              sys.exit(0 if success else 1)
          except Exception as e:
              logger.error(f"Unhandled script exception: {e}", exc_info=True)
              sys.exit(1)
      
      Output
      obb_test_jp5  | Creating new Ultralytics Settings v0.0.6 file ✅ 
      obb_test_jp5  | View Ultralytics Settings with 'yolo settings' or at '/root/.config/Ultralytics/settings.json'
      obb_test_jp5  | Update Settings with 'yolo settings key=value', i.e. 'yolo settings runs_dir=path/to/dir'. For help see https://docs.ultralytics.com/quickstart/#ultralytics-settings.
      obb_test_jp5  | 2026-05-14 12:51:35,392 - INFO - ============================================================
      obb_test_jp5  | 2026-05-14 12:51:35,392 - INFO - Environment Verification (yolo checks)
      obb_test_jp5  | 2026-05-14 12:51:35,392 - INFO - ============================================================
      obb_test_jp5  | 2026-05-14 12:51:50,554 - INFO - 
      obb_test_jp5  | Ultralytics 8.4.50 🚀 Python-3.8.10 torch-2.1.0a0+41361538.nv23.06 CUDA:0 (Xavier, 6833MiB)
      obb_test_jp5  | Setup complete ✅ (2 CPUs, 6.7 GB RAM, 91.7/107.6 GB disk)
      obb_test_jp5  | 
      obb_test_jp5  | OS                     Linux-5.10.216-tegra-aarch64-with-glibc2.29
      obb_test_jp5  | Environment            Docker
      obb_test_jp5  | Python                 3.8.10
      obb_test_jp5  | Install                git
      obb_test_jp5  | Path                   /ultralytics/ultralytics
      obb_test_jp5  | RAM                    6.67 GB
      obb_test_jp5  | Disk                   91.7/107.6 GB
      obb_test_jp5  | CPU                    ARMv8 Processor rev 0 (v8l)
      obb_test_jp5  | CPU count              2
      obb_test_jp5  | GPU                    Xavier, 6833MiB
      obb_test_jp5  | GPU count              1
      obb_test_jp5  | CUDA                   11.4
      obb_test_jp5  | 
      obb_test_jp5  | numpy                  ✅ 1.23.5>=1.23.0
      obb_test_jp5  | matplotlib             ✅ 3.7.5>=3.3.0
      obb_test_jp5  | opencv-python-headless ✅ 4.13.0.92>=4.6.0
      obb_test_jp5  | pillow                 ✅ 10.4.0>=7.1.2
      obb_test_jp5  | pyyaml                 ✅ 6.0.3>=5.3.1
      obb_test_jp5  | requests               ✅ 2.32.4>=2.23.0
      obb_test_jp5  | scipy                  ✅ 1.10.1>=1.4.1
      obb_test_jp5  | torch                  ✅ 2.1.0a0+41361538.nv23.6>=1.8.0
      obb_test_jp5  | torchvision            ✅ 0.16.2+c6f3977>=0.9.0
      obb_test_jp5  | psutil                 ✅ 7.2.2>=5.8.0
      obb_test_jp5  | polars                 ✅ 1.8.2>=0.20.0
      obb_test_jp5  | ultralytics-thop       ✅ 2.0.19>=2.0.18
      obb_test_jp5  | torch                  ✅ 2.1.0a0+41361538.nv23.6!=2.4.0,>=1.8.0; sys_platform == "win32"
      obb_test_jp5  | 2026-05-14 12:51:50,562 - INFO - ============================================================
      obb_test_jp5  | 2026-05-14 12:51:50,563 - INFO - Minimal OBB Test - Export & Validation
      obb_test_jp5  | 2026-05-14 12:51:50,564 - INFO - ============================================================
      obb_test_jp5  | 2026-05-14 12:51:50,565 - INFO - 
      obb_test_jp5  | Running Ultralytics native export for: my-yolo26m-model.pt
      obb_test_jp5  | 2026-05-14 12:51:51,353 - ERROR - Ultralytics native export failed: ERROR ❌️ /app/models/my-yolo26m-model.pt requires numpy>=1.26.1, however numpy==1.23.5 is installed.
      obb_test_jp5  | Traceback (most recent call last):
      obb_test_jp5  |   File "/ultralytics/ultralytics/nn/tasks.py", line 1466, in torch_safe_load
      obb_test_jp5  |     ckpt = _load()
      obb_test_jp5  |   File "/ultralytics/ultralytics/nn/tasks.py", line 1463, in _load
      obb_test_jp5  |     return torch_load(file, map_location="cpu")
      obb_test_jp5  |   File "/ultralytics/ultralytics/utils/patches.py", line 181, in torch_load
      obb_test_jp5  |     return torch.load(*args, **kwargs)
      obb_test_jp5  |   File "/usr/local/lib/python3.8/dist-packages/torch/serialization.py", line 877, in load
      obb_test_jp5  |     return _load(opened_zipfile, map_location, pickle_module, **pickle_load_args)
      obb_test_jp5  |   File "/usr/local/lib/python3.8/dist-packages/torch/serialization.py", line 1253, in _load
      obb_test_jp5  |     result = unpickler.load()
      obb_test_jp5  |   File "/usr/local/lib/python3.8/dist-packages/torch/serialization.py", line 1246, in find_class
      obb_test_jp5  |     return super().find_class(mod_name, name)
      obb_test_jp5  | ModuleNotFoundError: No module named 'numpy._core'
      obb_test_jp5  | 
      obb_test_jp5  | The above exception was the direct cause of the following exception:
      obb_test_jp5  | 
      obb_test_jp5  | Traceback (most recent call last):
      obb_test_jp5  |   File "src/minimal.py", line 58, in test_export_to_engine
      obb_test_jp5  |     model = YOLO(str(pt_path))
      obb_test_jp5  |   File "/ultralytics/ultralytics/models/yolo/model.py", line 76, in __init__
      obb_test_jp5  |     super().__init__(model=model, task=task, verbose=verbose)
      obb_test_jp5  |   File "/ultralytics/ultralytics/engine/model.py", line 144, in __init__
      obb_test_jp5  |     self._load(model, task=task)
      obb_test_jp5  |   File "/ultralytics/ultralytics/engine/model.py", line 283, in _load
      obb_test_jp5  |     self.model, self.ckpt = load_checkpoint(weights)
      obb_test_jp5  |   File "/ultralytics/ultralytics/nn/tasks.py", line 1537, in load_checkpoint
      obb_test_jp5  |     ckpt, weight = torch_safe_load(weight)  # load ckpt
      obb_test_jp5  |   File "/ultralytics/ultralytics/nn/tasks.py", line 1489, in torch_safe_load
      obb_test_jp5  |     raise ModuleNotFoundError(
      obb_test_jp5  | ModuleNotFoundError: ERROR ❌️ /app/models/my-yolo26m-model.pt requires numpy>=1.26.1, however numpy==1.23.5 is installed.
      obb_test_jp5  | 2026-05-14 12:51:51,382 - INFO - 
      obb_test_jp5  | Running trtexec to compile ONNX: my-yolo26m-model.onnx
      obb_test_jp5  | 2026-05-14 12:51:51,382 - INFO - Running command: /usr/src/tensorrt/bin/trtexec --onnx=/app/models/my-yolo26m-model.onnx --saveEngine=/app/models/my-yolo26m-model.trt.engine --fp16 --memPoolSize=workspace:768
      obb_test_jp5  | 2026-05-14 13:21:39,214 - INFO - trtexec successfully saved engine to /app/models/my-yolo26m-model.trt.engine
      obb_test_jp5  | 2026-05-14 13:21:39,217 - INFO - 
      obb_test_jp5  | Validating inference with engine: /app/models/my-yolo26m-model.trt.engine
      obb_test_jp5  | Ultralytics 8.4.50 🚀 Python-3.8.10 torch-2.1.0a0+41361538.nv23.06 CUDA:0 (Xavier, 6833MiB)
      obb_test_jp5  | Loading /app/models/my-yolo26m-model.trt.engine for TensorRT inference...
      obb_test_jp5  | [05/14/2026-13:21:41] [TRT] [I] Loaded engine size: 43 MiB
      obb_test_jp5  | [05/14/2026-13:21:45] [TRT] [I] [MemUsageChange] Init cuBLAS/cuBLASLt: CPU +260, GPU +306, now: CPU 648, GPU 5277 (MiB)
      obb_test_jp5  | [05/14/2026-13:21:46] [TRT] [I] [MemUsageChange] Init cuDNN: CPU +82, GPU +156, now: CPU 730, GPU 5433 (MiB)
      obb_test_jp5  | [05/14/2026-13:21:46] [TRT] [I] [MemUsageChange] TensorRT-managed allocation in engine deserialization: CPU +0, GPU +41, now: CPU 0, GPU 41 (MiB)
      obb_test_jp5  | [05/14/2026-13:21:46] [TRT] [I] [MemUsageChange] Init cuBLAS/cuBLASLt: CPU +0, GPU +28, now: CPU 687, GPU 5422 (MiB)
      obb_test_jp5  | [05/14/2026-13:21:46] [TRT] [I] [MemUsageChange] Init cuDNN: CPU +0, GPU +32, now: CPU 687, GPU 5454 (MiB)
      obb_test_jp5  | [05/14/2026-13:21:46] [TRT] [I] [MemUsageChange] TensorRT-managed allocation in IExecutionContext creation: CPU +0, GPU +92, now: CPU 0, GPU 133 (MiB)
      obb_test_jp5  | Setting batch=1 input of shape (1, 3, 1024, 1024)
      obb_test_jp5  | 2026-05-14 13:21:54,562 - INFO - generated new fontManager
      obb_test_jp5  | val: Fast image access ✅ (ping: 1.1±0.9 ms, read: 479.2±146.1 MB/s, size: 2563.4 KB)
      val: Scanning /app/data/labels/rec26.cache... 2940 images, 26 backgrounds, 0 corrupt: 100% ━━━━━━━━━━━━ 2940/2940 14.4Mit/s 0.0s
      obb_test_jp5  | 2026-05-14 13:22:01,033 - ERROR - Validation failed on /app/models/my-yolo26m-model.trt.engine: CUDA error: no kernel image is available for execution on the device
      obb_test_jp5  | CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
      obb_test_jp5  | For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
      obb_test_jp5  | Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.
      obb_test_jp5  | Traceback (most recent call last):
      obb_test_jp5  |   File "src/minimal.py", line 114, in test_export_to_engine
      obb_test_jp5  |     metrics = model.val(data=dataset_yaml, imgsz=1024, batch=1, verbose=False)
      obb_test_jp5  |   File "/ultralytics/ultralytics/engine/model.py", line 615, in val
      obb_test_jp5  |     validator(model=self.model)
      obb_test_jp5  |   File "/usr/local/lib/python3.8/dist-packages/torch/utils/_contextlib.py", line 115, in decorate_context
      obb_test_jp5  |     return func(*args, **kwargs)
      obb_test_jp5  |   File "/ultralytics/ultralytics/engine/validator.py", line 210, in __call__
      obb_test_jp5  |     model.warmup(imgsz=(1 if pt else self.args.batch, self.data["channels"], imgsz, imgsz))  # warmup
      obb_test_jp5  |   File "/ultralytics/ultralytics/nn/autobackend.py", line 313, in warmup
      obb_test_jp5  |     non_max_suppression(warmup_boxes)  # warmup NMS
      obb_test_jp5  |   File "/ultralytics/ultralytics/utils/nms.py", line 154, in non_max_suppression
      obb_test_jp5  |     i = torchvision.ops.nms(boxes, scores, iou_thres)
      obb_test_jp5  |   File "/usr/local/lib/python3.8/dist-packages/torchvision/ops/boxes.py", line 41, in nms
      obb_test_jp5  |     return torch.ops.torchvision.nms(boxes, scores, iou_threshold)
      obb_test_jp5  |   File "/usr/local/lib/python3.8/dist-packages/torch/_ops.py", line 646, in __call__
      obb_test_jp5  |     return self._op(*args, **kwargs or {})
      obb_test_jp5  | RuntimeError: CUDA error: no kernel image is available for execution on the device
      obb_test_jp5  | CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
      obb_test_jp5  | For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
      obb_test_jp5  | Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.
      obb_test_jp5  | 
      obb_test_jp5  | 2026-05-14 13:22:01,521 - INFO - 
      obb_test_jp5  | Execution complete.
      obb_test_jp5  | 2026-05-14 13:22:01,524 - INFO - 
      obb_test_jp5  | ============================================================
      obb_test_jp5  | 2026-05-14 13:22:01,526 - INFO - ✗ Test failed
      obb_test_jp5  | 2026-05-14 13:22:01,529 - INFO - ============================================================
      obb_test_jp5 exited with code 1

      Export Device

      This is a bit of info on where my ONNX files come from and what my PC environment looks like where I compiled them.

      Yolo Checks on the ONNX Export Device
      Ultralytics 8.4.48 🚀 Python-3.13.9 torch-2.11.0+cu130 CPU (AMD Ryzen 5 5600G with Radeon Graphics)
      Setup complete ✅ (12 CPUs, 15.0 GB RAM, 744.1/878.1 GB disk)
      
      OS                     Linux-6.17.0-23-generic-x86_64-with-glibc2.39
      Environment            Linux
      Python                 3.13.9
      Install                git
      Path                   /home/myuser/projects/myproject/.venv/lib/python3.13/site-packages/ultralytics
      RAM                    15.01 GB
      Disk                   744.1/878.1 GB
      CPU                    AMD Ryzen 5 5600G with Radeon Graphics
      CPU count              12
      GPU                    None
      GPU count              None
      CUDA                   None
      
      numpy                  ✅ 2.4.4>=1.23.0
      matplotlib             ✅ 3.10.9>=3.3.0
      opencv-python          ✅ 4.13.0.92>=4.6.0
      pillow                 ✅ 12.2.0>=7.1.2
      pyyaml                 ✅ 6.0.3>=5.3.1
      requests               ✅ 2.33.1>=2.23.0
      scipy                  ✅ 1.17.1>=1.4.1
      torch                  ✅ 2.11.0>=1.8.0
      torch                  ✅ 2.11.0!=2.4.0,>=1.8.0; sys_platform == "win32"
      torchvision            ✅ 0.26.0>=0.9.0
      psutil                 ✅ 7.2.2>=5.8.0
      polars                 ✅ 1.40.1>=0.20.0
      ultralytics-thop       ✅ 2.0.19>=2.0.18
      Export Code
      from ultralytics import YOLO
      
      # 1. Load the model weights
      model_path = "my-yolo26m-model.pt"
      model = YOLO(model_path)
      
      # 2. Export the model
      # Configured for OBB (Oriented Bounding Boxes) and legacy TensorRT compatibility
      success = model.export(
          format="onnx", 
          imgsz=1024,      # Target resolution
          opset=13,       # ONNX operator set version, also tried 12
          simplify=True,   # I hoped this would help
          task="obb",
          half=False,      # I tried this with True and False
          end2end=False    # For legacy TensorRT support...?
      )
      
      print(f"Export successful: {success}")
      YOLO26m-obb summary (fused): 175 layers, 23,443,236 parameters, 0 gradients, 71.5 GFLOPs
      
      PyTorch: starting from 'my-yolo26m-model.pt' with input shape (1, 3, 1024, 1024) BCHW and output shape(s) (1, 6, 21504) (90.8 MB)
      
      ONNX: starting export with onnx 1.21.0 opset 13...
      ONNX: slimming with onnxslim 0.1.93...
      ONNX: export success ✅ 2.5s, saved as 'my-yolo26m-model.onnx' (81.3 MB)
      
      Export complete (4.3s)

      Old Container

      Jut in case it is relevant, this is the yolo checks of my self-made container which is based on l4t-ml.

      root@xaviernx-desktop:/app# yolo checks
      WARNING ⚠️ user config directory '/root/.config/Ultralytics' is not writable, using '/tmp/Ultralytics'. Set YOLO_CONFIG_DIR to override.
      Creating new Ultralytics Settings v0.0.6 file ✅ 
      View Ultralytics Settings with 'yolo settings' or at '/tmp/Ultralytics/settings.json'
      Update Settings with 'yolo settings key=value', i.e. 'yolo settings runs_dir=path/to/dir'. For help see https://docs.ultralytics.com/quickstart/#ultralytics-settings.
      Ultralytics 8.4.49 🚀 Python-3.8.10 torch-2.0.0+nv23.05 CUDA:0 (Xavier, 6833MiB)
      Setup complete ✅ (2 CPUs, 6.7 GB RAM, 84.9/107.6 GB disk)
      
      OS                     Linux-5.10.216-tegra-aarch64-with-glibc2.29
      Environment            Docker
      Python                 3.8.10
      Install                pip
      Path                   /usr/local/lib/python3.8/dist-packages/ultralytics
      RAM                    6.67 GB
      Disk                   84.9/107.6 GB
      CPU                    ARMv8 Processor rev 0 (v8l)
      CPU count              2
      GPU                    Xavier, 6833MiB
      GPU count              1
      CUDA                   11.4
      
      numpy                  ✅ 1.23.5>=1.23.0
      matplotlib             ✅ 3.7.5>=3.3.0
      opencv-python          ❌ (not installed)>=4.6.0
      pillow                 ✅ 10.2.0>=7.1.2
      pyyaml                 ✅ 5.3.1>=5.3.1
      requests               ✅ 2.32.4>=2.23.0
      scipy                  ✅ 1.10.1>=1.4.1
      torch                  ✅ 2.0.0+nv23.5>=1.8.0
      torch                  ✅ 2.0.0+nv23.5!=2.4.0,>=1.8.0; sys_platform == "win32"
      torchvision            ✅ 0.15.1a0+42759b1>=0.9.0
      psutil                 ✅ 5.9.6>=5.8.0
      polars                 ❌ (not installed)>=0.20.0
      ultralytics-thop       ❌ (not installed)>=2.0.18
    3. Murat Raimbekov

      Thanks for the detailed MRE. I reproduced the numpy failure end-to-end with torch.save/torch.load and have a verified fix. There are two independent issues here.

      1. numpy._core / "requires numpy>=1.26.1" — root cause and fix

      Your .pt was saved on your PC with numpy 2.x / Python 3.13, so the checkpoint pickle embeds the numpy-2 numpy._core module path. JetPack 5 on Xavier NX runs Python 3.8, whose maximum numpy is 1.24.x, and Xavier cannot move to JetPack 6 (Orin-only). I verified that even numpy 1.26.x does not fully read such a checkpoint, so "upgrade numpy" is not a real option on JP5.

      The reliable fix is a small compatibility shim that aliases numpy._core to the legacy numpy.core, injected before the model is loaded. I confirmed this loads a numpy-2 torch.save checkpoint correctly under numpy 1.23.5 with the model and all fields intact. Put this at the very top of your main.py, before from ultralytics import YOLO:

      import sys, numpy, numpy.core
      sys.modules["numpy._core"] = numpy.core
      for _s in ("multiarray", "numeric", "umath", "_multiarray_umath"):
          try:
              sys.modules[f"numpy._core.{_s}"] = __import__(f"numpy.core.{_s}", fromlist=[_s])
          except Exception:
              pass

      Alternatively, fix the checkpoint once on the PC (where numpy 2.x can read it) so the device never needs the shim — load it and re-save it after converting the auxiliary numpy fields, then copy the new .pt to the Jetson.

      2. Empty boxes / CUDA error: no kernel image is available for execution on the device

      These are separate from numpy:

      • The trtexec-built engine produces no detections because a raw trtexec engine carries none of the Ultralytics export metadata, so the OBB decoding/post-processing is never applied. Build the engine on the Jetson with Ultralytics directly from the .pt (after applying the shim above) so metadata is embedded: YOLO("model.pt").export(format="engine", imgsz=1024, half=True, device=0).
      • no kernel image is available is raised inside torchvision.ops.nms on the GPU: the PyTorch/torchvision in ultralytics/ultralytics:latest-jetson-jetpack5 is built for Orin (SM 8.7), while Xavier NX is SM 7.2. That is exactly why YOLO11 OBB works in your older l4t-ml-based container (torch 2.0.0+nv23.5) but the official image fails. Recommended path: stay on the container that already works on your Xavier, pip install ultralytics opencv-python-headless there, and run the Ultralytics export(format="engine") + val() entirely on-device.

      TL;DR: the numpy error is a pickle module-path mismatch from the numpy-2 export PC (fix with the shim above — verified), and the empty/CUDA errors are a raw-trtexec metadata loss plus a torch/torchvision build that doesn't target Xavier's SM 7.2.

      1
    4. RotatingGiraffeGitHub

      Hi there, thank you for the help! I just recently got around to trying out the shim and now everything works just fine, thank you for your help!

  7. yashpatel2010GitHub

    How to run deepstream for YOLO26 models in my jetson orin nano developer kit?

    Is there any documentation for deepstream on jetson? I don't want any installation errors of version missmatch.

    So give me any perfect guide link so that I will not miss anything while installation and run inference on live RTSP streams

    1. Glenn JocherUltralytics

      Adding one detail for Jetson Orin Nano: keep the versions paired as JetPack 6.1 + DeepStream 7.1 from the Jetson quick-start and NVIDIA DeepStream getting started; for live RTSP, set the uri in deepstream_app_config.txt to your stream, and if you hit any mismatch please post the full text traceback plus yolo checks output here so the team can help.

    2. yashpatel2010GitHub

      Thanks for your reply. Just onemore doubt to clear. docker pull nvcr.io/nvidia/deepstream:7.1-triton-multiarch

      if I use above docker container for my jetpack6.1 so it there is no issue of version missmatch of deepstream right?

      is it works very well with Yolo26 deepstream settings which is showing in this doc? https://docs.ultralytics.com/guides/deepstream-nvidia-jetson/

      what is your recommendation to run Multiple RTSP streams with deepstream with docker containers or without docker containers manual installation?

    3. Glenn JocherUltralytics

      Yes, nvcr.io/nvidia/deepstream:7.1-triton-multiarch is the correct DeepStream 7.1 Jetson container for JetPack 6.1, but it still depends on the host having the JetPack BSP and NVIDIA container runtime installed, so it lowers mismatch risk rather than removing it completely; for YOLO26 + RTSP, the DeepStream multi-stream section is the right reference, and I’d suggest native setup first since that’s the install path used in the guide, then move to Docker once your pipeline is stable. (docs.ultralytics.com)

    4. yashpatel2010GitHub(modificato)

      Thanks. For native setup Is it good to install deepstream 7.1 while installing jetpack6.1 only?

  8. githubwaltGitHub

    Installing torch 2.11.0 from the official website pytorch.org on an AGX Orin 64Gb (dev kit) with JP6.2.2 or installing torch 2.10.0 and torchvision 0.25.0 following this page, give complaints about not being compatible with the Compute Capability (CC) of my machine which has 8.7, whatever that means. Only installing the previous versions torch 2.8.0 and torchvision 0.23.0 do work.

    I understand that torch needs to be tailored specifically for the Jetson family and that is what this page is all about, so why do I have this issue if I follow the instructions given here? Is it because I am on Jetpack 6.2.2 and here it's 6.1?

    1. githubwaltGitHub

      I forgot the warning and error output when running a python test script:

      UserWarning: Found GPU0 Orin which is of compute capability (CC) 8.7.The following list shows the CCs this version of PyTorch was built for and the hardware CCs it supports:- 8.0 which supports hardware CC >=8.0,<9.0 except {8.7}- 9.0 which supports hardware CC >=9.0,<10.0

      torch.AcceleratorError CUDA error: no kernel image is available for execution on the device

    2. Glenn JocherUltralytics

      Thanks, that traceback helps — CC 8.7 is the AGX Orin GPU architecture, and no kernel image is available means the installed PyTorch wheel was built without sm_87 support. So yes, this is most likely a JetPack/PyTorch wheel mismatch: the native steps in the JetPack 6.1 section don’t necessarily map cleanly to JP6.2.2, even though the page includes JP6.2 benchmark coverage. For now I’d stick with the working torch 2.8.0 + torchvision 0.23.0 pair or use the JetPack 6 Docker setup; if you want us to check your exact environment, please post yolo checks and the full text traceback here in the discussion.

    3. Glenn JocherUltralytics

      No worries — the duplicate is fine, and if you later confirm a stable JP6.2.2 + PyTorch pairing on AGX Orin, please add it here so we can reflect it in the Jetson guide.

    4. githubwaltGitHub

      Will do. Now trying to get the most out of the AGX. Experimenting with number of workers on a batch size of 16. So far 12 workers gave memory issues ... I chose 12 because of 12 cores, but that was a little too much. Now running with 8 workers ... so far so good. To be continued 🙂

    5. githubwaltGitHub

      Hi Glenn, I can confirm a stable JP6.2.2 + Pytorch (torch 2.8.0 and torchvision 0.23.0) on the AGX Orin.