Commencer

Utilisation de Python

Ce guide explique comment intégrer Ultralytics YOLO à tes projets Python pour la détection d’objets, la segmentation d’instances, la segmentation sémantique, l’estimation de la profondeur, la classification, l’estimation de pose et la détection par boîtes englobantes orientées. Tu apprendras à charger et utiliser des modèles préentraînés, à entraîner de nouveaux modèles et à effectuer des prédictions sur des images. L’interface Python te permet d’ajouter YOLO à un projet avec seulement quelques lignes de code.



Regarder : Maîtriser Ultralytics YOLO : Python

Par exemple, les utilisateurs peuvent charger un modèle, l’entraîner, évaluer ses performances sur un jeu de validation et même l’exporter au format ONNX en quelques lignes de code.

Python
from ultralytics import YOLO

# Créer un nouveau modèle YOLO à partir de zéro
model = YOLO("yolo26n.yaml")

# Charger un modèle YOLO préentraîné (recommandé pour l’entraînement)
model = YOLO("yolo26n.pt")

# Entraîner le modèle avec le jeu de données 'coco8.yaml' pendant 3 époques
results = model.train(data="coco8.yaml", epochs=3)

# Évaluer les performances du modèle sur le jeu de validation
results = model.val()

# Effectuer une détection d’objets dans une image à l’aide du modèle
results = model("https://ultralytics.com/images/bus.jpg")

# Exporter le modèle au format ONNX
success = model.export(format="onnx")

Entraînement#

Le mode Train sert à entraîner un modèle YOLO sur un jeu de données personnalisé. Dans ce mode, le modèle est entraîné à l’aide du jeu de données et des hyperparamètres spécifiés. Le processus d’entraînement consiste à optimiser les paramètres du modèle pour qu’il prédise avec précision les classes et les emplacements des objets dans une image.

Entraînement
from ultralytics import YOLO

model = YOLO("yolo26n.pt")  # transmettre n’importe quel type de modèle
results = model.train(epochs=5)

Exemples d’entraînement

Validation#

Le mode Val sert à valider un modèle YOLO après son entraînement. Dans ce mode, le modèle est évalué sur un jeu de validation afin de mesurer son exactitude et ses performances de généralisation. Ce mode peut servir à régler les hyperparamètres du modèle afin d’améliorer ses performances.

Validation
from ultralytics import YOLO

# Charger un modèle YOLO
model = YOLO("yolo26n.yaml")

# Entraîner le modèle
model.train(data="coco8.yaml", epochs=5)

# Valider sur la partition de validation du jeu de données d’entraînement
model.val()

Exemples de validation

Prédiction#

Le mode Predict sert à effectuer des prédictions sur de nouvelles images ou vidéos à l’aide d’un modèle YOLO entraîné. Dans ce mode, le modèle est chargé depuis un fichier de checkpoint, et tu peux fournir des images ou des vidéos pour effectuer l’inférence. Le modèle prédit les classes et les emplacements des objets dans les images ou vidéos d’entrée.

Prédiction
import cv2
from PIL import Image

from ultralytics import YOLO

model = YOLO("model.pt")
# accepte tous les formats - image/dir/Path/URL/video/PIL/ndarray. 0 pour la webcam
results = model.predict(source="folder", show=True)  # Afficher les prédictions. Accepte tous les arguments YOLO predict

# from PIL
im1 = Image.open("bus.jpg")
results = model.predict(source=im1, save=True)  # enregistrer les images tracées

# from ndarray
im2 = cv2.imread("bus.jpg")
results = model.predict(source=im2, save=True, save_txt=True)  # enregistrer les prédictions sous forme d’étiquettes

# à partir d’une liste de PIL/ndarray
results = model.predict(source=[im1, im2])

Exemples de prédiction

Export#

Le mode Export sert à exporter un modèle YOLO dans un format adapté au déploiement. Dans ce mode, le modèle est converti dans un format utilisable par d’autres applications logicielles ou appareils matériels. Ce mode est utile pour déployer le modèle dans des environnements de production.

Export

Exporte un modèle YOLO officiel au format ONNX avec une taille de lot et une taille d’image dynamiques.

from ultralytics import YOLO

model = YOLO("yolo26n.pt")
model.export(format="onnx", dynamic=True)

Exemples d’exportation

Suivi#

Le mode Track sert à suivre des objets en temps réel à l’aide d’un modèle YOLO. Dans ce mode, le modèle est chargé depuis un fichier de point de contrôle, et l’utilisateur peut fournir un flux vidéo en direct pour effectuer le suivi d’objets en temps réel. Ce mode est utile pour des applications telles que les systèmes de surveillance ou les voitures autonomes.

Suivi
from ultralytics import YOLO

# Charger un modèle
model = YOLO("yolo26n.pt")  # charger un modèle de détection officiel
model = YOLO("yolo26n-seg.pt")  # charger un modèle de segmentation officiel
model = YOLO("path/to/best.pt")  # charger un modèle personnalisé

# Effectuer le suivi avec le modèle
results = model.track(source="https://youtu.be/LNwODJXcvt4", show=True)
results = model.track(source="https://youtu.be/LNwODJXcvt4", show=True, tracker="bytetrack.yaml")

Exemples de suivi

Benchmark#

Le mode Benchmark sert à évaluer la vitesse et la précision de différents formats d’exportation pour YOLO. Les benchmarks fournissent des informations sur la taille du format exporté, sa métrique de tâche (mAP50-95 pour la détection, la segmentation, la pose et OBB, mIoU pour la segmentation sémantique, δ1 pour l’estimation de profondeur ou accuracy_top1 pour la classification), ainsi que le temps d’inférence en millisecondes par image pour différents formats d’exportation tels que ONNX, OpenVINO, TensorRT et d’autres. Ces informations peuvent aider à choisir le format d’exportation optimal selon les exigences de vitesse et de précision de ton cas d’utilisation.

Benchmark

Évaluer un modèle YOLO officiel dans tous les formats d’exportation.

from ultralytics.utils.benchmarks import benchmark

# Benchmark
benchmark(model="yolo26n.pt", data="coco8.yaml", imgsz=640, device=0)

Exemples de benchmark

Utiliser les entraîneurs#

La classe de modèle YOLO sert d’interface de haut niveau aux classes Trainer. Chaque tâche YOLO possède son propre entraîneur, qui hérite de BaseTrainer. Cette architecture offre davantage de flexibilité et de possibilités de personnalisation dans tes workflows de machine learning.

Exemple d’entraîneur de détection
from ultralytics.models.yolo.detect import DetectionPredictor, DetectionTrainer, DetectionValidator

overrides = {"model": "yolo26n.pt", "data": "coco8.yaml", "epochs": 3}

# Trainer
trainer = DetectionTrainer(overrides=overrides)
trainer.train()
trained_model = trainer.best  # chemin vers best.pt

# Validator
validator = DetectionValidator(args={"data": "coco8.yaml"})
validator(model=trained_model)

# Predictor
predictor = DetectionPredictor(overrides={"model": trained_model})
predictor(source="https://ultralytics.com/images/bus.jpg")

# Reprendre un entraînement interrompu à partir de son dernier checkpoint
trainer = DetectionTrainer(overrides={"resume": "path/to/last.pt"})
trainer.train()

Tu peux personnaliser les Trainers pour prendre en charge des tâches personnalisées ou explorer des idées de recherche et développement. La conception modulaire d’Ultralytics YOLO te permet d’adapter le framework à tes besoins spécifiques, que tu travailles sur une nouvelle tâche de vision par ordinateur ou que tu ajustes des modèles existants pour améliorer leurs performances.

Tutoriels de personnalisation

FAQ#

  • L’intégration d’Ultralytics YOLO à tes projets Python est simple. Tu peux charger un modèle préentraîné ou entraîner un nouveau modèle à partir de zéro. Voici comment commencer :

    from ultralytics import YOLO
    
    # Charger un modèle YOLO préentraîné
    model = YOLO("yolo26n.pt")
    
    # Effectuer la détection d’objets sur une image
    results = model("https://ultralytics.com/images/bus.jpg")
    
    # Visualiser les résultats
    for result in results:
        result.show()

    Consulte des exemples plus détaillés dans la section Mode Predict.

  • Ultralytics YOLO propose six modes pour différents flux de travail d’apprentissage automatique :

    • Train : entraîne un modèle à l’aide de jeux de données personnalisés.
    • Val : valide les performances du modèle sur un ensemble de validation.
    • Predict : effectue des prédictions sur de nouvelles images ou des flux vidéo.
    • Export : exporte des modèles dans différents formats, comme ONNX et TensorRT.
    • Track : effectue le suivi d’objets en temps réel dans des flux vidéo.
    • Benchmark : évalue les performances du modèle avec différentes configurations.

    Chaque mode couvre une étape différente du développement et du déploiement de modèles.

  • Pour entraîner un modèle YOLO personnalisé, indique ton jeu de données et les autres hyperparamètres. Voici un exemple rapide :

    from ultralytics import YOLO
    
    # Charger le modèle YOLO
    model = YOLO("yolo26n.yaml")
    
    # Entraîner le modèle avec un jeu de données personnalisé
    model.train(data="path/to/your/dataset.yaml", epochs=10)
  • Avec la fonction export, exporter des modèles YOLO dans un format adapté au déploiement est simple. Par exemple, tu peux exporter un modèle au format ONNX :

    from ultralytics import YOLO
    
    # Charger le modèle YOLO
    model = YOLO("yolo26n.pt")
    
    # Exporter le modèle au format ONNX
    model.export(format="onnx")

    Pour connaître les différentes options d’exportation, consulte la documentation du Mode Export.

  • Oui. Après l’entraînement, transmets un autre fichier YAML de jeu de données à val() pour y évaluer le modèle :

    from ultralytics import YOLO
    
    # Charger un modèle YOLO
    model = YOLO("yolo26n.yaml")
    
    # Entraîner le modèle
    model.train(data="coco8.yaml", epochs=5)
    
    # Valider le modèle sur un autre jeu de données
    model.val(data="path/to/separate/data.yaml")

    Consulte la page Mode Val pour des exemples détaillés et des informations sur son utilisation.

Commentaires31

Pose une question, partage tes résultats ou aide quelqu’un d’autre.

  1. kindaichinisanGitHub

    How do I detect object using onnx model exported from pt model (yolo v9). Any sample code? Is it required to do pre-processing and post-processing?

    1. Glenn JocherUltralytics

      @kindaichinisan you can use ONNX models directly with Ultralytics for inference - no manual pre/post-processing needed:

      from ultralytics import YOLO
      
      # Load the ONNX model
      model = YOLO("your_model.onnx")
      
      # Run inference
      results = model("image.jpg")

      The Ultralytics framework handles all pre-processing and post-processing automatically when using ONNX models, just like with PT models.

  2. UltralyticsAssistantGitHubUltralytics

    👋 Hello, thank you for your interest in Ultralytics 🚀! We recommend checking out the Docs for new users, particularly the Python Usage guide you’ve linked, which contains detailed instructions on integrating YOLO for object detection, segmentation, and classification.

    If this is a 🐛 Bug Report, please provide a minimum reproducible example so we can assist you better.

    For custom training or ❓ questions, please include as much detail as possible, such as your Python script, dataset examples, training logs, or errors, and ensure you’re following our Tips for Best Training Results.

    Looking to connect with the Ultralytics community? Join us:

    • For real-time chat, visit Discord 🎧
    • For focused discussions, try Discourse
    • For general discussions and tips, explore our Subreddit

    Upgrade

    Ensure you’re running the latest version of the ultralytics package with all its requirements. Upgrade in a Python>=3.8 environment with PyTorch>=1.8 by running:

    pip install -U ultralytics

    Environments

    YOLO works seamlessly in any of the following verified environments (preloaded with dependencies like CUDA, Python, and PyTorch):

    Status

    Ultralytics CI

    If the badge above is green, our Ultralytics CI tests are passing. These tests are run daily and verify correct operation of all YOLO Modes and Tasks across macOS, Windows, and Ubuntu.

    This is an automated response to assist you quickly. An Ultralytics engineer will review your question and provide additional support shortly 😊

    1. dnnxlGitHub

      @dnnxl yes, YOLO11 supports augmentations during training, which are defined in the dataset configuration YAML file. You can customize augmentations such as flipping, scaling, and cropping to enhance model performance. For more details, refer to the Train Mode documentation. Let us know if you have further questions!

      Thanks for your answer, it works for object detection purposes right?

    2. Glenn JocherUltralytics

      Yes, augmentations in YOLO11 are fully supported for object detection tasks and can be applied during training to improve model generalization. Let us know if you need guidance on configuring them!

  3. Rain020GitHub

    Operating system: Win11, Ultralyicts: 8.3.27, YOLO11. I have developed a desktop application for PC-side processing. I need to release related GPU memory resources after closing the page or completing tasks. However, I haven’t found an effective way to release the memory to prevent GPU memory leaks. Looking forward to your response.

    1. Glenn JocherUltralytics

      @Rain020 to release GPU memory after completing tasks with YOLO11, you can try using PyTorch's torch.cuda.empty_cache() function to clear unused cached memory. Additionally, ensure that all model and variable references are deleted and use del to remove them before calling the cache clearing function. If issues persist, consider checking for any persistent references that might still occupy memory. For more details, refer to the PyTorch documentation.

    2. Rain020GitHub

      Thank you for your response. I need to conduct some experiments to verify whether our solution is effective.

  4. Sheepy1977GitHub

    There are some translation-mistakes in Chinese "Train" should be "训练",not “火车” "Val" should be "校验",not "瓦尔" "Track“ should be "跟踪“,not "轨道"

    1. Paula DerrengerUltralytics

      @Sheepy1977 thank you for pointing out the translation errors. We appreciate your attention to detail. We will update the translations as follows:

      • "Train" will be corrected to "训练"
      • "Val" will be corrected to "校验"
      • "Track" will be corrected to "跟踪"

      Your feedback helps us improve the documentation for all users. If you notice any other issues, please let us know.

  5. yirenyixinGitHub

    Hello, when I was training with the dataset, I found that all the label names I gave were changed to class_X after training. The following is the configuration of my ymal file: path: E:\BaiduNetdiskDownload\Citypersons train: images/train val: images/train nc: 1 name: 0: person

    After training, use the model to recognize others. The label information displays class_0

    1. Glenn JocherUltralytics

      Hello! It seems like the class names are defaulting to class_X because the model might not be recognizing the class names specified in your YAML file. This can sometimes happen if there's an issue with the formatting of the YAML file or how the classes are being read during training.

      To ensure your class names are recognized correctly, please make sure your YAML file is formatted correctly. Here's a corrected version of your YAML configuration:

      path: E:\BaiduNetdiskDownload\Citypersons
      train: images/train
      val: images/train
      nc: 1
      names:
        - person

      Notice the change from name: to names: and ensuring that the class names are listed with a dash -. This should help the model recognize and use the correct class names during training and inference.

      If the issue persists after this adjustment, please double-check the path and file names for typos and ensure that the dataset is being loaded correctly during training. If you need further assistance, feel free to ask! 😊

  6. flyingship19820207GitHub

    The code is simple, for it is the same with your sample:

    from ultralytics import YOLO

    Create a new YOLO model from scratch

    model = YOLO('yolov8n.yaml')

    model = YOLO('yolov8n.pt')

    Train the model using the 'coco128.yaml' dataset for 3 epochs

    results = model.train(data='coco128.yaml', epochs=3)

    Evaluate the model's performance on the validation set

    results = model.val()

    Perform object detection on an image using the model

    results = model('https://ultralytics.com/images/bus.jpg')

    Export the model to ONNX format

    success = model.export(format='onnx')

    Below is the error details: Traceback (most recent call last): File "/usr/local/lib/python3.11/site-packages/ultralytics/engine/trainer.py", line 133, in init self.data = check_det_dataset(self.args.data) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.11/site-packages/ultralytics/data/utils.py", line 271, in check_det_dataset if zipfile.is_zipfile(file) or is_tarfile(file): ^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.11/tarfile.py", line 2791, in is_tarfile t = open(name) ^^^^^^^^^^ File "/usr/local/lib/python3.11/tarfile.py", line 1804, in open return func(name, "r", fileobj, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.11/tarfile.py", line 1933, in xzopen t = cls.taropen(name, mode, fileobj, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.11/tarfile.py", line 1854, in taropen return cls(name, mode, fileobj, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.11/tarfile.py", line 1714, in init self.firstmember = self.next() ^^^^^^^^^^^ File "/usr/local/lib/python3.11/tarfile.py", line 2631, in next raise e File "/usr/local/lib/python3.11/tarfile.py", line 2604, in next tarinfo = self.tarinfo.fromtarfile(self) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.11/tarfile.py", line 1292, in fromtarfile buf = tarfile.fileobj.read(BLOCKSIZE) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.11/lzma.py", line 207, in read return self._buffer.read(size) ^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.11/_compression.py", line 68, in readinto data = self.read(len(byte_view)) ^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.11/_compression.py", line 96, in read if self._decompressor.needs_input: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ AttributeError: '_lzma.LZMADecompressor' object has no attribute 'needs_input'

    The above exception was the direct cause of the following exception:

    Traceback (most recent call last): File "/root/chenyong/1_yolo/1_test/yolo_test.py", line 10, in results = model.train(data='coco128.yaml', epochs=3) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.11/site-packages/ultralytics/engine/model.py", line 635, in train self.trainer = (trainer or self._smart_load("trainer"))(overrides=args, _callbacks=self.callbacks) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.11/site-packages/ultralytics/engine/trainer.py", line 137, in init raise RuntimeError(emojis(f"Dataset '{clean_url(self.args.data)}' error ❌ {e}")) from e RuntimeError: Dataset 'coco128.yaml' error ❌ '_lzma.LZMADecompressor' object has no attribute 'needs_input'

    1. Paula DerrengerUltralytics

      It seems like you're encountering an issue related to the LZMA compression library, specifically with the _lzma.LZMADecompressor object. This could be due to a variety of reasons, including compatibility issues with the Python version you're using or the specific dataset file format.

      As a first step, I recommend ensuring that your Python environment is up-to-date and that you're using a compatible version of the Ultralytics YOLO package. Additionally, verifying the integrity and format of the coco128.yaml dataset file might help, as the error suggests there could be an issue with reading the dataset file.

      If you're still facing this issue after checking these points, could you please provide more details about your environment, such as the Python and Ultralytics YOLO versions? This information will help us better understand the problem and assist you more effectively. 🛠️

      In the meantime, here's a quick check you can perform to ensure your dataset file is accessible and correctly formatted:

      import yaml
      
      # Attempt to load the YAML file to check its integrity
      with open('coco128.yaml', 'r') as file:
          dataset = yaml.safe_load(file)
          print(dataset)

      This code snippet will try to load your coco128.yaml file and print its contents, which can help identify if there are any issues with the file itself.

    2. flyingship19820207GitHub

      Thanks for your reply. I copy coco128.yaml from /usr/local/lib/python3.11/site-packages/ultralytics/cfg/datasets/coco128.yaml to current work dir.And run the code you provided: import yaml

      Attempt to load the YAML file to check its integrity

      with open('coco128.yaml', 'r') as file: dataset = yaml.safe_load(file) print(dataset) The result is as below: {'path': '../datasets/coco128', 'train': 'images/train2017', 'val': 'images/train2017', 'test': None, 'names': {0: 'person', 1: 'bicycle', 2: 'car', 3: 'motorcycle', 4: 'airplane', 5: 'bus', 6: 'train', 7: 'truck', 8: 'boat', 9: 'traffic light', 10: 'fire hydrant', 11: 'stop sign', 12: 'parking meter', 13: 'bench', 14: 'bird', 15: 'cat', 16: 'dog', 17: 'horse', 18: 'sheep', 19: 'cow', 20: 'elephant', 21: 'bear', 22: 'zebra', 23: 'giraffe', 24: 'backpack', 25: 'umbrella', 26: 'handbag', 27: 'tie', 28: 'suitcase', 29: 'frisbee', 30: 'skis', 31: 'snowboard', 32: 'sports ball', 33: 'kite', 34: 'baseball bat', 35: 'baseball glove', 36: 'skateboard', 37: 'surfboard', 38: 'tennis racket', 39: 'bottle', 40: 'wine glass', 41: 'cup', 42: 'fork', 43: 'knife', 44: 'spoon', 45: 'bowl', 46: 'banana', 47: 'apple', 48: 'sandwich', 49: 'orange', 50: 'broccoli', 51: 'carrot', 52: 'hot dog', 53: 'pizza', 54: 'donut', 55: 'cake', 56: 'chair', 57: 'couch', 58: 'potted plant', 59: 'bed', 60: 'dining table', 61: 'toilet', 62: 'tv', 63: 'laptop', 64: 'mouse', 65: 'remote', 66: 'keyboard', 67: 'cell phone', 68: 'microwave', 69: 'oven', 70: 'toaster', 71: 'sink', 72: 'refrigerator', 73: 'book', 74: 'clock', 75: 'vase', 76: 'scissors', 77: 'teddy bear', 78: 'hair drier', 79: 'toothbrush'}, 'download': 'https://ultralytics.com/assets/coco128.zip'}

    3. Paula DerrengerUltralytics(modifié)

      @flyingship19820207 hey there! Thanks for reaching out with your environment details. It looks like you're using Python 3.11.7 and YOLOv8 version 8.1.24, which should be compatible.

      Given the error you encountered, it seems like there might be an issue with the LZMA library in Python 3.11.7 or how the dataset is being processed. As a quick workaround, you might want to try running your code in a virtual environment with an earlier version of Python (e.g., Python 3.8 or 3.9) to see if the issue persists. This can help isolate whether it's a Python version-specific issue.

      Here's a quick guide on setting up a virtual environment:

      # Create a virtual environment with Python 3.8 (needs to be installed already)
      python3.8 -m venv yolov8-env
      
      # Activate the virtual environment
      source yolov8-env/bin/activate
      
      # Install Ultralytics YOLO
      pip install ultralytics

      After setting up and activating your virtual environment, try running your YOLOv8 code again. If the problem is resolved, it might be worth reporting this as a compatibility issue with Python 3.11.7 on the Ultralytics GitHub page for further investigation. 🕵️‍♂️

      Let me know how it goes!

    4. flyingship19820207GitHub

      I have tried python3.10 which system-provided by ubuntu 22.4,it works well.It seems the error is caused by many python version in one system.

    5. Paula DerrengerUltralytics

      Hey there! 🎉 It's great to hear that switching to Python 3.10 resolved the issue for you. It seems like the problem was indeed related to having multiple Python versions in your system. This can sometimes lead to conflicts, especially with dependencies and libraries that are sensitive to the Python version.

      For future reference, it's always a good idea to use virtual environments for each project. This way, you can manage dependencies more effectively and avoid version conflicts. Here's a quick way to set up a virtual environment for your projects:

      python3 -m venv myenv
      source myenv/bin/activate

      And then, you can install the specific packages you need within this isolated environment.

      If you have any more questions or run into further issues, feel free to reach out. Happy coding with YOLOv8! 🚀

  7. flyingship19820207GitHub

    there is an error while trying to train yolo8 :RuntimeError: Dataset 'coco8.yaml' error ❌ '_lzma.LZMADecompressor' object has no attribute 'needs_input'.I use python3.11 and the system is ubuntu.How to fix it.

    1. Paula DerrengerUltralytics

      Hey there! 👋 It seems like you're encountering an issue related to the _lzma module while training YOLOv8 with Python 3.11 on Ubuntu. This could be due to a compatibility issue with the LZMA library in Python 3.11.

      A quick workaround could be to try downgrading to Python 3.8 or 3.9, as these versions are known to work well with YOLOv8 and its dependencies. You can manage multiple Python versions using a tool like pyenv.

      Here's how you can install and use pyenv to switch to Python 3.8:

      # Install pyenv
      curl https://pyenv.run | bash
      
      # Add pyenv to path (you might want to add this to your .bashrc or .zshrc)
      export PATH="$HOME/.pyenv/bin:$PATH"
      eval "$(pyenv init --path)"
      eval "$(pyenv virtualenv-init -)"
      
      # Install Python 3.8
      pyenv install 3.8.10
      
      # Set the local (current directory) Python version to 3.8.10
      pyenv local 3.8.10
      
      # Verify the Python version
      python --version

      After switching to Python 3.8, try running your training command again. If the issue persists or you have any more questions, feel free to reach out! 😊

    2. flyingship19820207GitHub

      I comment these code below: #if zipfile.is_zipfile(file) or is_tarfile(file):

      new_dir = safe_download(file, dir=DATASETS_DIR, unzip=True, delete=False)

      file = find_dataset_yaml(DATASETS_DIR / new_dir)

      extract_dir, autodownload = file.parent, False

      in file /usr/local/lib/python3.11/site-packages/ultralytics/data/utils.py,then run again, it seems works.

    3. Paula DerrengerUltralytics(modifié)

      Hey there! 👋 It looks like you've found a workaround by commenting out some code related to dataset extraction in utils.py. While this might work temporarily, it's generally not recommended to modify library code directly, as it could lead to unexpected behavior or issues with future updates.

      The error you encountered seems related to the _lzma module in Python 3.11. This could be a compatibility issue with the Python version and the LZMA library. A safer approach might be to ensure all your dependencies are correctly installed and compatible with Python 3.11. You could try reinstalling the lzma module or using a Python environment with a version known to work well with YOLOv8, such as Python 3.8 or 3.9.

      If you're set on using Python 3.11, you might also want to check if there's an open issue on the Ultralytics YOLOv8 GitHub repository regarding this, or consider opening a new issue to bring this to the developers' attention. They might offer a more permanent fix or advice on handling this compatibility issue.

      In the meantime, if you need to keep working, using a virtual environment with a different Python version could be a quick fix:

      # Create a virtual environment with Python 3.8
      python3.8 -m venv yolov8-env
      
      # Activate the virtual environment
      source yolov8-env/bin/activate
      
      # Install YOLOv8 and other dependencies
      pip install ultralytics

      Hope this helps, and happy coding! 😊