xView-Datensatz
Der xView-Datensatz ist einer der größten öffentlich verfügbaren Benchmarks für Satellitenbilder zur Objekterkennung. Er umfasst mehr als eine Million Objektinstanzen aus 60 Klassen, die mit Begrenzungsrahmen in über 1.400 km² WorldView-3-Bildmaterial mit einer Auflösung von 0,3 m annotiert sind. Er wurde für die DIUx-xView-2018-Challenge von der National Geospatial-Intelligence Agency (NGA) der USA veröffentlicht und muss manuell heruntergeladen werden; der Download ist etwa 20,7 GB groß.
Der Datensatz wurde erstellt, um vier Bereiche der Computer Vision voranzubringen:
- Die Mindestauflösung für die Erkennung verringern.
- Die Lerneffizienz verbessern.
- Die Entdeckung weiterer Objektklassen ermöglichen.
- Die Erkennung feingranularer Klassen verbessern.
Aufbauend auf Benchmarks wie COCO konzentriert sich xView auf Luft- und Satellitenbilder, in denen Objekte viel kleiner und dichter angeordnet sind als auf Bodenaufnahmen.
Der xView-Datensatz wird nicht automatisch heruntergeladen. Registriere dich auf der Website der DIUx xView Challenge 2018, um train_images.zip (~15 GB), train_labels.zip und val_images.zip (~5 GB) herunterzuladen. Entpacke die Dateien anschließend unter datasets/xView/, sodass folgendes Verzeichnis entsteht:
datasets/xView/
├── train_images/ # 847 TIF satellite images
├── val_images/ # 282 TIF images (no public labels)
└── xView_train.geojson # bounding-box annotationsBeim ersten Trainingslauf konvertiert Ultralytics die GeoJSON-Annotationen ins YOLO-Format und teilt die annotierten Bilder automatisch ungefähr im Verhältnis 90/10 in Trainings- und Validierungsdaten auf – eine manuelle Konvertierung ist nicht erforderlich.
Wichtige Funktionen#
- Feingranulare Klassen: 60 Objektklassen aus den Bereichen Luftfahrzeuge, Fahrzeuge, Schienenfahrzeuge, Schiffe, Baumaschinen und Gebäude – viele davon klein, selten und visuell ähnlich.
- Hohe Auflösung: Bodenauflösung von 0,3 m, erfasst von WorldView-3-Satelliten.
- Dichte Annotationen: Über 1 Million Objektinstanzen auf mehr als 1.400 km² Bildmaterial, alle mit horizontalen Begrenzungsrahmen gekennzeichnet.
- Automatische Konvertierung: Das Ultralytics-Downloadskript konvertiert die ursprünglichen GeoJSON-Beschriftungen ins YOLO-Format und erstellt bei der ersten Verwendung die Aufteilung in Trainings- und Validierungsdaten.
Datensatzstruktur#
xView-Bilder sind großflächige Satellitenaufnahmen im TIF-Format. Nur für die 847 Trainingsbilder sind öffentliche Beschriftungen verfügbar – für den Validierungssatz der Challenge mit 282 Bildern gibt es keine. Daher teilt die Ultralytics-Konfiguration xView.yaml die annotierten Bilder bei der ersten Verwendung automatisch auf:
| Aufteilung | Bilder | Beschreibung |
|---|---|---|
| Trainieren | ~90 % von 847 | In autosplit_train.txt aufgeführte annotierte Bilder, die beim ersten Lauf generiert werden |
| Validierung | ~10 % von 847 | In autosplit_val.txt aufgeführte annotierte Bilder, die zur Auswertung verwendet werden |
Die 60 Klassen umfassen feingranulare Kategorien wie Starrflügelflugzeug, Frachtflugzeug, Kleinwagen, Bus, Lokomotive, Schiff, Bagger, Gebäude, Flugzeughangar und Lagertank. Die vollständige Liste findest du in der Datensatz-YAML-Datei unten. Bei der Konvertierung werden die ursprünglichen Klassen-IDs der Challenge (11–94) den fortlaufenden Indizes 0–59 zugeordnet.
Anwendungsfälle#
Dank seiner feingranularen Klassen und der hochauflösenden Aufsichtsperspektive ist xView ein Standardbenchmark zum Trainieren und Bewerten von Deep-Learning-Modellen für die Fernerkundung. Häufige Anwendungsbereiche sind:
- Militärische Aufklärung und Verteidigung
- Stadtplanung und Stadtentwicklung
- Umweltüberwachung
- Katastrophenschutz und Schadensbewertung
- Kartierung und Verwaltung von Infrastruktur
Weitere Benchmarks für Luft- und Satellitenbilder findest du im auf Drohnen spezialisierten VisDrone-Datensatz oder im DOTA-v2-Datensatz mit orientierten Begrenzungsrahmen.
YAML-Datei des Datensatzes#
Die Datei xView.yaml definiert die Datensatzkonfiguration: die Datensatzpfade, die 60 Klassennamen und das Downloadskript, das die GeoJSON-Annotationen konvertiert und die automatische Aufteilung erstellt. Die Datei wird im Ultralytics-Repository unter https://github.com/ultralytics/ultralytics/blob/main/ultralytics/cfg/datasets/xView.yaml gepflegt.
# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
# DIUx xView 2018 Challenge dataset https://challenge.xviewdataset.org by U.S. National Geospatial-Intelligence Agency (NGA)
# -------- Download and extract data manually to `datasets/xView` before running the train command. --------
# Documentation: https://docs.ultralytics.com/datasets/detect/xview
# Example usage: yolo train data=xView.yaml
# parent
# ├── ultralytics
# └── datasets
# └── xView ← downloads here (20.7 GB)
# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
path: xView # dataset root dir
train: images/autosplit_train.txt # train images (relative to 'path') 90% of 847 train images
val: images/autosplit_val.txt # val images (relative to 'path') 10% of 847 train images
# Classes
names:
0: Fixed-wing Aircraft
1: Small Aircraft
2: Cargo Plane
3: Helicopter
4: Passenger Vehicle
5: Small Car
6: Bus
7: Pickup Truck
8: Utility Truck
9: Truck
10: Cargo Truck
11: Truck w/Box
12: Truck Tractor
13: Trailer
14: Truck w/Flatbed
15: Truck w/Liquid
16: Crane Truck
17: Railway Vehicle
18: Passenger Car
19: Cargo Car
20: Flat Car
21: Tank car
22: Locomotive
23: Maritime Vessel
24: Motorboat
25: Sailboat
26: Tugboat
27: Barge
28: Fishing Vessel
29: Ferry
30: Yacht
31: Container Ship
32: Oil Tanker
33: Engineering Vehicle
34: Tower crane
35: Container Crane
36: Reach Stacker
37: Straddle Carrier
38: Mobile Crane
39: Dump Truck
40: Haul Truck
41: Scraper/Tractor
42: Front loader/Bulldozer
43: Excavator
44: Cement Mixer
45: Ground Grader
46: Hut/Tent
47: Shed
48: Building
49: Aircraft Hangar
50: Damaged Building
51: Facility
52: Construction Site
53: Vehicle Lot
54: Helipad
55: Storage Tank
56: Shipping container lot
57: Shipping Container
58: Pylon
59: Tower
# Download script/URL (optional) ---------------------------------------------------------------------------------------
download: |
import json
from pathlib import Path
import shutil
import numpy as np
from PIL import Image
from ultralytics.utils import TQDM
from ultralytics.data.split import autosplit
from ultralytics.utils.ops import xyxy2xywhn
def convert_labels(fname=Path("xView/xView_train.geojson")):
"""Convert xView GeoJSON labels to YOLO format (classes 0-59) and save them as text files."""
path = fname.parent
with open(fname, encoding="utf-8") as f:
print(f"Loading {fname}...")
data = json.load(f)
# Make dirs
labels = path / "labels" / "train"
shutil.rmtree(labels, ignore_errors=True)
labels.mkdir(parents=True, exist_ok=True)
# xView classes 11-94 to 0-59
xview_class2index = [-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 0, 1, 2, -1, 3, -1, 4, 5, 6, 7, 8, -1, 9, 10, 11,
12, 13, 14, 15, -1, -1, 16, 17, 18, 19, 20, 21, 22, -1, 23, 24, 25, -1, 26, 27, -1, 28, -1,
29, 30, 31, 32, 33, 34, 35, 36, 37, -1, 38, 39, 40, 41, 42, 43, 44, 45, -1, -1, -1, -1, 46,
47, 48, 49, -1, 50, 51, -1, 52, -1, -1, -1, 53, 54, -1, 55, -1, -1, 56, -1, 57, -1, 58, 59]
shapes = {}
for feature in TQDM(data["features"], desc=f"Converting {fname}"):
p = feature["properties"]
if p["bounds_imcoords"]:
image_id = p["image_id"]
image_file = path / "train_images" / image_id
if image_file.exists(): # 1395.tif missing
try:
box = np.array([int(num) for num in p["bounds_imcoords"].split(",")])
assert box.shape[0] == 4, f"incorrect box shape {box.shape[0]}"
cls = p["type_id"]
cls = xview_class2index[int(cls)] # xView class to 0-59
assert 59 >= cls >= 0, f"incorrect class index {cls}"
# Write YOLO label
if image_id not in shapes:
shapes[image_id] = Image.open(image_file).size
box = xyxy2xywhn(box[None].astype(float), w=shapes[image_id][0], h=shapes[image_id][1], clip=True)
with open((labels / image_id).with_suffix(".txt"), "a", encoding="utf-8") as f:
f.write(f"{cls} {' '.join(f'{x:.6f}' for x in box[0])}\n") # write label.txt
except Exception as e:
print(f"WARNING: skipping one label for {image_file}: {e}")
# Download manually from https://challenge.xviewdataset.org
dir = Path(yaml["path"]) # dataset root dir
# urls = [
# "https://d307kc0mrhucc3.cloudfront.net/train_labels.zip", # train labels
# "https://d307kc0mrhucc3.cloudfront.net/train_images.zip", # 15G, 847 train images
# "https://d307kc0mrhucc3.cloudfront.net/val_images.zip", # 5G, 282 val images (no labels)
# ]
# download(urls, dir=dir)
# Convert labels
convert_labels(dir / "xView_train.geojson")
# Move images
images = Path(dir / "images")
images.mkdir(parents=True, exist_ok=True)
Path(dir / "train_images").rename(dir / "images" / "train")
Path(dir / "val_images").rename(dir / "images" / "val")
# Split
autosplit(dir / "images" / "train")Verwendung#
Zum Trainieren müssen die oben beschriebenen manuell heruntergeladenen Dateien unter datasets/xView/ entpackt werden. Anschließend werden die Annotationen automatisch konvertiert und die Trainings- und Validierungsdaten aufgeteilt.
Um ein Modell 100 Epochen lang mit Bildgröße 640 auf dem xView-Datensatz zu trainieren, kannst du die folgenden Codeausschnitte verwenden.
from ultralytics import YOLO
# Ein Modell laden
model = YOLO("yolo26n.pt") # ein vortrainiertes Modell laden (für das Training empfohlen)
# Das Modell trainieren
results = model.train(data="xView.yaml", epochs=100, imgsz=640)Um zusätzliche Satellitenbilder zu annotieren und xView-Trainingsläufe im Browser zu verwalten, verwende die Ultralytics Platform.
Beispieldaten und Annotationen#
Das folgende Beispiel zeigt eine typische xView-Szene: hochauflösende Luft- und Satellitenbilder, auf denen kleine Objekte wie Fahrzeuge und Gebäude mit Begrenzungsrahmen annotiert sind. Das veranschaulicht, warum die Objekterkennung auf Satellitenbildern eine feingranulare Lokalisierung erfordert.

Zitate und Danksagungen#
Wenn du den xView-Datensatz in deiner Forschung oder Entwicklungsarbeit verwendest, zitiere bitte die folgende Arbeit:
@misc{lam2018xview,
title={xView: Objects in Context in Overhead Imagery},
author={Darius Lam and Richard Kuzma and Kevin McGee and Samuel Dooley and Michael Laielli and Matthew Klaric and Yaroslav Bulatov and Brendan McCord},
year={2018},
eprint={1802.07856},
archivePrefix={arXiv},
primaryClass={cs.CV}
}Wir möchten der Einheit für Verteidigungsinnovation (DIU) und den Erstellern des xView-Datensatzes für ihren wertvollen Beitrag zur Computer-Vision-Forschungsgemeinschaft danken. Weitere Informationen findest du auf der xView-Datensatz-Website.
Häufig gestellte Fragen#
Der xView-Datensatz ist ein Benchmark für Satellitenbilder, der für die DIUx xView Challenge 2018 von der US-amerikanischen Nationalen Behörde für Geoinformation und Aufklärung veröffentlicht wurde. Er umfasst über 1 Million Objektinstanzen aus 60 feingranularen Klassen in WorldView-3-Bildmaterial mit einer Auflösung von 0,3 m. Der Datensatz unterstützt die Forschung zur Erkennung kleiner, seltener und feingranularer Objekte aus der Vogelperspektive, die wesentlich schwieriger zu erkennen sind als Objekte auf Bodenaufnahmen.
xView muss manuell heruntergeladen werden: Registriere dich auf der Website der DIUx xView Challenge 2018, lade
train_images.zip(~15 GB),train_labels.zipundval_images.zip(~5 GB) herunter – insgesamt etwa 20,7 GB – und entpacke sie entsprechend der oben auf dieser Seite in der Warnung gezeigten Verzeichnisstruktur unterdatasets/xView/. Beim ersten Trainingslauf konvertiert Ultralytics die GeoJSON-Annotationen automatisch ins YOLO-Format und erstellt die Aufteilung in Trainings- und Validierungsdaten.xView enthält 847 annotierte Trainingsbilder und 282 Validierungsbilder ohne öffentliche Beschriftungen. Alle wurden von WorldView-3-Satelliten mit einer Auflösung von 0,3 m aufgenommen. Die Annotationen umfassen über 1 Million Objektinstanzen aus 60 Klassen. Da nur die Trainingsbeschriftungen öffentlich verfügbar sind, teilt die Ultralytics-Konfiguration
xView.yamldie 847 annotierten Bilder ungefähr im Verhältnis 90/10 in Trainings- und Validierungsdaten auf. Weitere Informationen findest du unter Datensatzstruktur.Trainiere ein YOLO26n-Modell 100 Epochen lang mit xView und einer Bildgröße von 640:
Trainingsbeispielfrom ultralytics import YOLO # Ein Modell laden model = YOLO("yolo26n.pt") # ein vortrainiertes Modell laden (für das Training empfohlen) # Das Modell trainieren results = model.train(data="xView.yaml", epochs=100, imgsz=640)Ausführliche Informationen zu Argumenten und Einstellungen findest du auf der Seite zum Training des Modells.
Zitiere die Arbeit „xView: Objects in Context in Overhead Imagery“ (Lam et al., arXiv:1802.07856, 2018). Den vollständigen BibTeX-Eintrag findest du im obigen Abschnitt Zitate und Danksagungen.
Kommentare25
Stell eine Frage, teile deine Ergebnisse oder hilf jemand anderem.
XiFromEY
Thank you for sharing the code @glenn-jocher and @jing-qiu. If I only want to detect building and damaged building, how should I modify the code above? Any hints? Thank you very much!
1XiFromEY
I figured out using ChatGPT. However, I am looking at the imbalance issue.
Glenn JocherUltralytics(bearbeitet)@XiFromEY hey there! 👋 To focus on detecting buildings and damaged buildings, you'll want to modify your dataset's classes. The simplest way would be to create a script to update your *.txt labels to delete any rows that don't pertain to the classes you are interested in, and then train normally with your updated YAML that points to your modified labels:
yolo detect train data=xView.yaml model=yolo26n.pt epochs=100 imgsz=640This will train your model to detect just the building classes you're interested in. Happy coding! 🚀
chandan-chakma
after download the images i can not convert properly to yolo-format . datasets/xView/ images/train_images /val_images xView_train_geojson could you please share the conversion code?
Glenn JocherUltralyticsThe conversion code is already bundled in the
download:section of the dataset YAML—see the maintainedxView.yamlconversion script (also referenced from the xView dataset docs); place your manual download asdatasets/xView/train_images/,datasets/xView/val_images/, anddatasets/xView/xView_train.geojson(not underimages/), then run:yolo detect train data=xView.yaml model=yolo26n.pt imgsz=640 epochs=1If it still fails, please paste the full text traceback here (not a screenshot) and your
yolo checksoutput so we can pinpoint what’s off in the paths/GeoJSON.chandan-chakma
manual download as datasets/xView/train_images/, datasets/xView/val_images/, and datasets/xView/xView_train.geojson after run yolo detect train data=xView.yaml model=yolo26n.pt imgsz=640 epochs=1 can it formated or i need to do manually convert it?
chandan-chakma
i am getting this error. (.new) (base) alpha@alpha-HP-Z2-Tower-G9-Workstation-Desktop-PC:~/Desktop/chakma/test$ python train.py Ultralytics 8.4.7 🚀 Python-3.11.7 torch-2.9.1+cu128 CUDA:0 (NVIDIA GeForce RTX 3090, 24115MiB) engine/trainer: agnostic_nms=False, amp=True, angle=1.0, augment=False, auto_augment=randaugment, batch=16, bgr=0.0, box=7.5, cache=False, cfg=None, classes=None, close_mosaic=10, cls=0.5, compile=False, conf=None, copy_paste=0.0, copy_paste_mode=flip, cos_lr=False, cutmix=0.0, data=xView.yaml, degrees=0.0, deterministic=True, device=None, dfl=1.5, dnn=False, dropout=0.0, dynamic=False, embed=None, epochs=100, erasing=0.4, exist_ok=False, fliplr=0.5, flipud=0.0, format=torchscript, fraction=1.0, freeze=None, half=False, hsv_h=0.015, hsv_s=0.7, hsv_v=0.4, imgsz=640, int8=False, iou=0.7, keras=False, kobj=1.0, line_width=None, lr0=0.01, lrf=0.01, mask_ratio=4, max_det=300, mixup=0.0, mode=train, model=yolov8n.pt, momentum=0.937, mosaic=1.0, multi_scale=0.0, name=train16, nbs=64, nms=False, opset=None, optimize=False, optimizer=auto, overlap_mask=True, patience=100, perspective=0.0, plots=True, pose=12.0, pretrained=True, profile=False, project=None, rect=False, resume=False, retina_masks=False, rle=1.0, save=True, save_conf=False, save_crop=False, save_dir=/home/alpha/Desktop/chakma/test/runs/detect/train16, save_frames=False, save_json=False, save_period=-1, save_txt=False, scale=0.5, seed=0, shear=0.0, show=False, show_boxes=True, show_conf=True, show_labels=True, simplify=True, single_cls=False, source=None, split=val, stream_buffer=False, task=detect, time=None, tracker=botsort.yaml, translate=0.1, val=True, verbose=True, vid_stride=1, visualize=False, warmup_bias_lr=0.1, warmup_epochs=3.0, warmup_momentum=0.8, weight_decay=0.0005, workers=8, workspace=None
WARNING ⚠️ Dataset 'xView.yaml' images not found, missing path '/home/alpha/Desktop/chakma/test/datasets/xView/images/autosplit_val.txt' Traceback (most recent call last): File "/home/alpha/Desktop/chakma/test/.new/lib/python3.11/site-packages/ultralytics/engine/trainer.py", line 660, in get_dataset data = check_det_dataset(self.args.data) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/alpha/Desktop/chakma/test/.new/lib/python3.11/site-packages/ultralytics/data/utils.py", line 471, in check_det_dataset exec(s, {"yaml": data}) File "", line 9, in ImportError: cannot import name 'autosplit' from 'ultralytics.data.utils' (/home/alpha/Desktop/chakma/test/.new/lib/python3.11/site-packages/ultralytics/data/utils.py)
The above exception was the direct cause of the following exception:
Traceback (most recent call last): File "/home/alpha/Desktop/chakma/test/train.py", line 7, in results = model.train(data="xView.yaml", epochs=100, imgsz=640) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/alpha/Desktop/chakma/test/.new/lib/python3.11/site-packages/ultralytics/engine/model.py", line 769, in train self.trainer = (trainer or self._smart_load("trainer"))(overrides=args, _callbacks=self.callbacks) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/alpha/Desktop/chakma/test/.new/lib/python3.11/site-packages/ultralytics/models/yolo/detect/train.py", line 63, in init super().init(cfg, overrides, _callbacks) File "/home/alpha/Desktop/chakma/test/.new/lib/python3.11/site-packages/ultralytics/engine/trainer.py", line 186, in init self.data = self.get_dataset() ^^^^^^^^^^^^^^^^^^ File "/home/alpha/Desktop/chakma/test/.new/lib/python3.11/site-packages/ultralytics/engine/trainer.py", line 664, in get_dataset raise RuntimeError(emojis(f"Dataset '{clean_url(self.args.data)}' error ❌ {e}")) from e RuntimeError: Dataset 'xView.yaml' error ❌ cannot import name 'autosplit' from 'ultralytics.data.utils' (/home/alpha/Desktop/chakma/test/.new/lib/python3.11/site-packages/ultralytics/data/utils.py) (.new) (base) alpha@alpha-HP-Z2-Tower-G9-Workstation-Desktop-PC:~/Desktop/chakma/test$
Glenn JocherUltralyticsThat
ImportError: cannot import name 'autosplit'means you’re running a stalexView.yamldownload:script that importsautosplitfrom the wrong module (it’s whyimages/autosplit_val.txtnever gets created), so you don’t need to manually convert—just update Ultralytics and ensure you’re using thexView.yamlthat ships with your installed package (not an older local copy); the dataset conversion is triggered automatically bycheck_det_dataset()when files are missing (see the dataset check/download logic).pip install -U ultralytics yolo checks yolo detect train data=xView.yaml model=yolo26n.pt imgsz=640 epochs=1If it still fails, paste your
yolo checksoutput and the first ~15 lines of thedownload:block from the exactxView.yamlfile your run is loading (plus the full traceback).chandan-chakma
thanks finally it working. i trained on yolov8n the result is quite low. map50 and map50-95 and recall low but prec is high is it normal or have some issue.here is the script and the result.
import os os.environ['OPENCV_LOG_LEVEL'] = 'ERROR' import warnings warnings.filterwarnings('ignore') from ultralytics import YOLO
Load a model
model = YOLO("yolov8n.pt") # load a pretrained model (recommended for training)
Train the model
results = model.train(data="xView.yaml", epochs=100, imgsz=640, batch=4,amp=True,workers=4,cache=False)
====================
Validating /home/alpha/Desktop/chakma/test/runs/detect/train46/weights/best.pt... Ultralytics 8.4.7 🚀 Python-3.11.7 torch-2.9.1+cu128 CUDA:0 (NVIDIA GeForce RTX 3090, 24115MiB) Model summary (fused): 73 layers, 3,017,348 parameters, 0 gradients, 8.1 GFLOPs Class Images Instances Box(P R mAP50 mAP50-95): 18% ━━────────── 2/11 1.1s/it 0.7s<9.8sWARNING ⚠️ NMS time limit 2.400s exceeded Class Images Instances Box(P R mAP50 mAP50-95): 100% ━━━━━━━━━━━━ 11/11 1.2it/s 9.5s all 86 64690 0.61 0.00241 0.0044 0.00219 Fixed-wing Aircraft 4 5 0 0 0 0 Small Aircraft 7 144 0 0 0 0 Cargo Plane 16 127 1 0 0 0 Helicopter 5 10 1 0 0 0 Passenger Vehicle 22 114 1 0 0 0 Small Car 71 24345 0 0 0 0 Bus 45 645 1 0 0 0 Pickup Truck 25 156 1 0 0 0 Utility Truck 33 410 0 0 0 0 Truck 49 1536 1 0 0 0 Cargo Truck 48 666 1 0 0 0 Truck w/Box 29 343 0 0 0 0 Truck Tractor 17 39 1 0 0 0 Trailer 36 977 1 0 0 0 Truck w/Flatbed 21 85 1 0 0 0 Truck w/Liquid 5 9 0 0 0 0 Crane Truck 8 17 1 0 0 0 Railway Vehicle 1 2 1 0 0 0 Passenger Car 5 72 1 0 0 0 Cargo Car 4 23 0 0 0 0 Locomotive 1 1 1 0 0 0 Maritime Vessel 10 80 1 0 0 0 Motorboat 11 47 0 0 0 0 Sailboat 2 2 1 0 0 0 Tugboat 4 25 1 0 0 0 Barge 5 25 0 0 0 0 Fishing Vessel 11 115 0 0 0 0 Ferry 7 33 0 0 0 0 Yacht 4 7 1 0 0 0 Container Ship 5 31 1 0 0 0 Oil Tanker 5 9 0 0 0 0 Engineering Vehicle 13 27 1 0 0 0 Tower crane 8 18 0 0 0 0 Container Crane 1 1 0 0 0 0 Reach Stacker 3 8 1 0 0 0 Straddle Carrier 1 1 1 0 0 0 Mobile Crane 11 34 0 0 0 0 Dump Truck 29 120 1 0 0 0 Haul Truck 3 33 1 0 0 0 Scraper/Tractor 3 3 0 0 0 0 Front loader/Bulldozer 23 86 0 0 0 0 Excavator 15 97 1 0 0 0 Cement Mixer 7 11 1 0 0 0 Ground Grader 4 5 1 0 0 0 Hut/Tent 8 40 1 0 0 0 Shed 24 124 1 0 0 0 Building 79 32296 0.398 0.14 0.199 0.0823 Aircraft Hangar 3 13 0 0 0 0 Damaged Building 14 98 0 0 0 0 Facility 21 85 1 0 0 0 Construction Site 20 160 1 0 0 0 Vehicle Lot 39 597 1 0 0 0 Helipad 4 5 0 0 0 0 Storage Tank 23 132 0 0 0 0 Shipping container lot 15 404 1 0 0.0557 0.0446 Shipping Container 14 174 0 0 0 0 Pylon 7 15 1 0 0 0 Tower 3 3 1 0 0 0 Speed: 0.4ms preprocess, 9.2ms inference, 0.0ms loss, 66.1ms postprocess per image Results saved to /home/alpha/Desktop/chakma/test/runs/detect/train46
Glenn JocherUltralyticsYes—on xView this “high precision, near-zero recall/mAP” usually means the model is producing very few usable detections, and your
WARNING ⚠️ NMS time limit ... exceededduringvalis a strong hint the evaluation is getting truncated on these very dense scenes (lots of tiny objects per image), so the reported metrics can look artificially awful; the most effective fix is to train/validate on tiled crops (so each image has fewer objects) and/or increaseimgsz, and for a stronger baseline start fromyolo26n.pt/yolo26s.ptrather thanyolov8n.pt(see the relevant knobs in the Train mode docs).
chandan-chakma
WARN:0@12.381] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 33922 (0x8482) encountered [ WARN:0@12.381] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 34735 (0x87af) encountered [ WARN:0@12.381] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 34736 (0x87b0) encountered [ WARN:0@12.381] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 34737 (0x87b1) encountered [ WARN:0@12.381] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 42112 (0xa480) encountered [ WARN:0@12.381] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 42113 (0xa481) encountered Plotting labels to /home/alpha/Desktop/MH/ultralytics/runs/detect/train8/labels.jpg... optimizer: 'optimizer=auto' found, ignoring 'lr0=0.01' and 'momentum=0.937' and determining best 'optimizer', 'lr0' and 'momentum' automatically... optimizer: AdamW(lr=0.000156, momentum=0.9) with parameter groups 110 weight(decay=0.0), 173 weight(decay=0.0005), 184 bias(decay=0.0) Image sizes 1024 train, 1024 val Using 8 dataloader workers Logging results to /home/alpha/Desktop/MH/ultralytics/runs/detect/train8 Starting training for 100 epochs...
Epoch GPU_mem box_loss cls_loss dfl_loss Instances Size0%| | 0/191 [00:00<?, ?it/s][ WARN:0@12.764] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 33550 (0x830e) encountered [ WARN:0@12.764] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 33922 (0x8482) encountered [ WARN:0@12.764] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 34735 (0x87af) encountered [ WARN:0@12.764] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 34736 (0x87b0) encountered [ WARN:0@12.764] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 34737 (0x87b1) encountered [ WARN:0@12.764] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 42112 (0xa480) encountered [ WARN:0@12.764] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 42113 (0xa481) encountered [ WARN:0@12.802] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 33550 (0x830e) encountered [ WARN:0@12.802] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 33922 (0x8482) encountered [ WARN:0@12.802] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 34735 (0x87af) encountered [ WARN:0@12.802] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 34736 (0x87b0) encountered [ WARN:0@12.802] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 34737 (0x87b1) encountered [ WARN:0@12.802] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 42112 (0xa480) encountered [ WARN:0@12.802] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 42113 (0xa481) encountered [ WARN:0@12.842] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 33550 (0x830e) encountered [ WARN:0@12.842] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 33922 (0x8482) encountered [ WARN:0@12.842] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 34735 (0x87af) encountered [ WARN:0@12.842] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 34736 (0x87b0) encountered [ WARN:0@12.842] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 34737 (0x87b1) encountered [ WARN:0@12.842] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 42112 (0xa480) encountered [ WARN:0@12.842] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 42113 (0xa481) encountered [ WARN:0@12.869] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 33550 (0x830e) encountered [ WARN:0@12.869] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 33922 (0x8482) encountered [ WARN:0@12.869] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 34735 (0x87af) encountered [ WARN:0@12.869] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 34736 (0x87b0) encountered [ WARN:0@12.870] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 34737 (0x87b1) encountered [ WARN:0@12.870] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 42112 (0xa480) encountered [ WARN:0@12.870] global grfmt_tiff.cpp:122 TIFF_Warning TIFFReadDirectory: Unknown field with tag 42113 (0xa481) encountered 0%| | 0/191 [00:00<?, ?it/s] why i am getting this error could you please tell how can i solve it?
Glenn JocherUltralyticsThose are OpenCV/libtiff GeoTIFF metadata warnings from reading xView
.tiffiles (not a YOLO error), so if training continues you can ignore them; if your run actually stops, please paste the full text traceback plus the output of:yolo checksIf you just want to silence the warnings, you can run with:
OPENCV_LOG_LEVEL=ERROR yolo detect train data=xView.yaml model=yolo26n.pt imgsz=1024and if OpenCV can’t reliably read your TIFFs on your system, a practical workaround is to re-encode the
.tifto plain RGB (PNG/JPG) after confirming your folder layout matches the xView dataset guide.chandan-chakma
thank you very. so it will not effect model perfomance?
Glenn JocherUltralyticsYes—those GeoTIFF tag warnings are about metadata parsing and won’t affect model performance as long as the images decode correctly and training/validation completes; if you want to sanity-check decoding, run a quick inference on a few
.tiftiles (and if anything looks off, paste the full traceback +yolo checksoutput as in the Troubleshooting Common YOLO Issues guide).
Mahmoud-owais
New https://pypi.org/project/ultralytics/8.2.64 available 😃 Update with 'pip install -U ultralytics' Ultralytics YOLOv8.2.52 🚀 Python-3.10.12 torch-2.3.1+cu121 CUDA:0 (NVIDIA L4, 22700MiB) engine/trainer: task=detect, mode=train, model=yolov8n.pt, data=/content/drive/MyDrive/ultralytics-main/datasets/xView/xView.yaml, epochs=100, time=None, patience=100, batch=16, imgsz=640, save=True, save_period=-1, cache=False, device=None, workers=8, project=None, name=train63, exist_ok=False, pretrained=True, optimizer=auto, verbose=True, seed=0, deterministic=True, single_cls=False, rect=False, cos_lr=False, close_mosaic=10, resume=False, amp=True, fraction=1.0, profile=False, freeze=None, multi_scale=False, overlap_mask=True, mask_ratio=4, dropout=0.0, val=True, split=val, save_json=False, save_hybrid=False, conf=None, iou=0.7, max_det=300, half=False, dnn=False, plots=True, source=None, vid_stride=1, stream_buffer=False, visualize=False, augment=False, agnostic_nms=False, classes=None, retina_masks=False, embed=None, show=False, save_frames=False, save_txt=False, save_conf=False, save_crop=False, show_labels=True, show_conf=True, show_boxes=True, line_width=None, format=torchscript, keras=False, optimize=False, int8=False, dynamic=False, simplify=False, opset=None, workspace=4, nms=False, lr0=0.01, lrf=0.01, momentum=0.937, weight_decay=0.0005, warmup_epochs=3.0, warmup_momentum=0.8, warmup_bias_lr=0.1, box=7.5, cls=0.5, dfl=1.5, pose=12.0, kobj=1.0, label_smoothing=0.0, nbs=64, hsv_h=0.015, hsv_s=0.7, hsv_v=0.4, degrees=0.0, translate=0.1, scale=0.5, shear=0.0, perspective=0.0, flipud=0.0, fliplr=0.5, bgr=0.0, mosaic=1.0, mixup=0.0, copy_paste=0.0, auto_augment=randaugment, erasing=0.4, crop_fraction=1.0, cfg=None, tracker=botsort.yaml, save_dir=runs/detect/train63 Downloading https://ultralytics.com/assets/Arial.ttf to '/root/.config/Ultralytics/Arial.ttf'... 100%|██████████| 755k/755k [00:00<00:00, 123MB/s] Overriding model.yaml nc=80 with nc=60
from n params module arguments0 -1 1 464 ultralytics.nn.modules.conv.Conv [3, 16, 3, 2]
1 -1 1 4672 ultralytics.nn.modules.conv.Conv [16, 32, 3, 2]
2 -1 1 7360 ultralytics.nn.modules.block.C2f [32, 32, 1, True]
3 -1 1 18560 ultralytics.nn.modules.conv.Conv [32, 64, 3, 2]
4 -1 2 49664 ultralytics.nn.modules.block.C2f [64, 64, 2, True]
5 -1 1 73984 ultralytics.nn.modules.conv.Conv [64, 128, 3, 2]
6 -1 2 197632 ultralytics.nn.modules.block.C2f [128, 128, 2, True]
7 -1 1 295424 ultralytics.nn.modules.conv.Conv [128, 256, 3, 2]
8 -1 1 460288 ultralytics.nn.modules.block.C2f [256, 256, 1, True]
9 -1 1 164608 ultralytics.nn.modules.block.SPPF [256, 256, 5]
10 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
11 [-1, 6] 1 0 ultralytics.nn.modules.conv.Concat [1]
12 -1 1 148224 ultralytics.nn.modules.block.C2f [384, 128, 1]
13 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
14 [-1, 4] 1 0 ultralytics.nn.modules.conv.Concat [1]
15 -1 1 37248 ultralytics.nn.modules.block.C2f [192, 64, 1]
16 -1 1 36992 ultralytics.nn.modules.conv.Conv [64, 64, 3, 2]
17 [-1, 12] 1 0 ultralytics.nn.modules.conv.Concat [1]
18 -1 1 123648 ultralytics.nn.modules.block.C2f [192, 128, 1]
19 -1 1 147712 ultralytics.nn.modules.conv.Conv [128, 128, 3, 2]
20 [-1, 9] 1 0 ultralytics.nn.modules.conv.Concat [1]
21 -1 1 493056 ultralytics.nn.modules.block.C2f [384, 256, 1]
22 [15, 18, 21] 1 763012 ultralytics.nn.modules.head.Detect [60, [64, 128, 256]]
Model summary: 225 layers, 3022548 parameters, 3022532 gradients, 8.3 GFLOPsTransferred 319/355 items from pretrained weights TensorBoard: Start with 'tensorboard --logdir runs/detect/train63', view at http://localhost:6006/ Freezing layer 'model.22.dfl.conv.weight' AMP: running Automatic Mixed Precision (AMP) checks with YOLOv8n... AMP: checks passed ✅ train: Scanning /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train... 760 images, 0 backgrounds, 0 corrupt: 100%|██████████| 760/760 [02:31<00:00, 5.02it/s] train: New cache created: /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train.cache albumentations: Blur(p=0.01, blur_limit=(3, 7)), MedianBlur(p=0.01, blur_limit=(3, 7)), ToGray(p=0.01), CLAHE(p=0.01, clip_limit=(1, 4.0), tile_grid_size=(8, 8)) /usr/lib/python3.10/multiprocessing/popen_fork.py:66: RuntimeWarning: os.fork() was called. os.fork() is incompatible with multithreaded code, and JAX is multithreaded, so this will likely lead to a deadlock. self.pid = os.fork() val: Scanning /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train... 86 images, 0 backgrounds, 0 corrupt: 100%|██████████| 86/86 [00:16<00:00, 5.19it/s]val: New cache created: /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train.cache
Plotting labels to runs/detect/train63/labels.jpg... optimizer: 'optimizer=auto' found, ignoring 'lr0=0.01' and 'momentum=0.937' and determining best 'optimizer', 'lr0' and 'momentum' automatically... optimizer: AdamW(lr=0.000156, momentum=0.9) with parameter groups 57 weight(decay=0.0), 64 weight(decay=0.0005), 63 bias(decay=0.0) TensorBoard: model graph visualization added ✅ Image sizes 640 train, 640 val Using 8 dataloader workers Logging results to runs/detect/train63 Starting training for 100 epochs...
Epoch GPU_mem box_loss cls_loss dfl_loss Instances Size 1/100 18.3G 3.308 5.558 1.757 1.571e+04 640: 2%|▏ | 1/48 [00:05<04:33, 5.83s/it]
OutOfMemoryError Traceback (most recent call last) in <cell line: 7>() 5 6 # Train the model ----> 7 results = model.train(data="/content/drive/MyDrive/ultralytics-main/datasets/xView/xView.yaml", epochs=100, imgsz=640)
13 frames /content/drive/MyDrive/ultralytics-main/ultralytics/utils/tal.py in select_candidates_in_gts(xy_centers, gt_bboxes, eps) 225 bs, n_boxes, _ = gt_bboxes.shape 226 lt, rb = gt_bboxes.view(-1, 1, 4).chunk(2, 2) # left-top, right-bottom --> 227 bbox_deltas = torch.cat((xy_centers[None] - lt, rb - xy_centers[None]), dim=2).view(bs, n_boxes, n_anchors, -1) 228 # return (bbox_deltas.min(3)[0] > eps).to(gt_bboxes.dtype) 229 return bbox_deltas.amin(3).gt_(eps)
OutOfMemoryError: CUDA out of memory. Tried to allocate 8.63 GiB. GPU
Paula DerrengerUltralytics@Mahmoud-owais it looks like you're encountering an "Out of Memory" error during training. This can happen if the batch size or image size is too large for your GPU's memory. Try reducing the
batchsize orimgszto see if that resolves the issue. Additionally, ensure you're using the latest version of the Ultralytics package by runningpip install -U ultralytics. If the problem persists, please let us know. For more details on the xView dataset, you can visit xView Dataset Documentation.
Mahmoud-owais
Hello ! can help me What's the WARNING here ? New https://pypi.org/project/ultralytics/8.2.59 available 😃 Update with 'pip install -U ultralytics' Ultralytics YOLOv8.2.52 🚀 Python-3.10.12 torch-2.3.1+cu121 CPU (Intel Xeon 2.20GHz) engine/trainer: task=detect, mode=train, model=yolov8n.pt, data=xView.yaml, epochs=100, time=None, patience=100, batch=16, imgsz=640, save=True, save_period=-1, cache=False, device=None, workers=8, project=None, name=train8, exist_ok=False, pretrained=True, optimizer=auto, verbose=True, seed=0, deterministic=True, single_cls=False, rect=False, cos_lr=False, close_mosaic=10, resume=False, amp=True, fraction=1.0, profile=False, freeze=None, multi_scale=False, overlap_mask=True, mask_ratio=4, dropout=0.0, val=True, split=val, save_json=False, save_hybrid=False, conf=None, iou=0.7, max_det=300, half=False, dnn=False, plots=True, source=None, vid_stride=1, stream_buffer=False, visualize=False, augment=False, agnostic_nms=False, classes=None, retina_masks=False, embed=None, show=False, save_frames=False, save_txt=False, save_conf=False, save_crop=False, show_labels=True, show_conf=True, show_boxes=True, line_width=None, format=torchscript, keras=False, optimize=False, int8=False, dynamic=False, simplify=False, opset=None, workspace=4, nms=False, lr0=0.01, lrf=0.01, momentum=0.937, weight_decay=0.0005, warmup_epochs=3.0, warmup_momentum=0.8, warmup_bias_lr=0.1, box=7.5, cls=0.5, dfl=1.5, pose=12.0, kobj=1.0, label_smoothing=0.0, nbs=64, hsv_h=0.015, hsv_s=0.7, hsv_v=0.4, degrees=0.0, translate=0.1, scale=0.5, shear=0.0, perspective=0.0, flipud=0.0, fliplr=0.5, bgr=0.0, mosaic=1.0, mixup=0.0, copy_paste=0.0, auto_augment=randaugment, erasing=0.4, crop_fraction=1.0, cfg=None, tracker=botsort.yaml, save_dir=runs/detect/train8 Downloading https://ultralytics.com/assets/Arial.ttf to '/root/.config/Ultralytics/Arial.ttf'... 100%|██████████| 755k/755k [00:00<00:00, 15.5MB/s] Overriding model.yaml nc=80 with nc=60
from n params module arguments0 -1 1 464 ultralytics.nn.modules.conv.Conv [3, 16, 3, 2]
1 -1 1 4672 ultralytics.nn.modules.conv.Conv [16, 32, 3, 2]
2 -1 1 7360 ultralytics.nn.modules.block.C2f [32, 32, 1, True]
3 -1 1 18560 ultralytics.nn.modules.conv.Conv [32, 64, 3, 2]
4 -1 2 49664 ultralytics.nn.modules.block.C2f [64, 64, 2, True]
5 -1 1 73984 ultralytics.nn.modules.conv.Conv [64, 128, 3, 2]
6 -1 2 197632 ultralytics.nn.modules.block.C2f [128, 128, 2, True]
7 -1 1 295424 ultralytics.nn.modules.conv.Conv [128, 256, 3, 2]
8 -1 1 460288 ultralytics.nn.modules.block.C2f [256, 256, 1, True]
9 -1 1 164608 ultralytics.nn.modules.block.SPPF [256, 256, 5]
10 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
11 [-1, 6] 1 0 ultralytics.nn.modules.conv.Concat [1]
12 -1 1 148224 ultralytics.nn.modules.block.C2f [384, 128, 1]
13 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
14 [-1, 4] 1 0 ultralytics.nn.modules.conv.Concat [1]
15 -1 1 37248 ultralytics.nn.modules.block.C2f [192, 64, 1]
16 -1 1 36992 ultralytics.nn.modules.conv.Conv [64, 64, 3, 2]
17 [-1, 12] 1 0 ultralytics.nn.modules.conv.Concat [1]
18 -1 1 123648 ultralytics.nn.modules.block.C2f [192, 128, 1]
19 -1 1 147712 ultralytics.nn.modules.conv.Conv [128, 128, 3, 2]
20 [-1, 9] 1 0 ultralytics.nn.modules.conv.Concat [1]
21 -1 1 493056 ultralytics.nn.modules.block.C2f [384, 256, 1]
22 [15, 18, 21] 1 763012 ultralytics.nn.modules.head.Detect [60, [64, 128, 256]]
Model summary: 225 layers, 3022548 parameters, 3022532 gradientsTransferred 319/355 items from pretrained weights TensorBoard: Start with 'tensorboard --logdir runs/detect/train8', view at http://localhost:6006/ Freezing layer 'model.22.dfl.conv.weight' train: Scanning /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train/train_images... 0 images, 760 backgrounds, 0 corrupt: 100%|██████████| 760/760 [06:02<00:00, 2.09it/s]train: WARNING ⚠️ No labels found in /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train/train_images.cache. See https://docs.ultralytics.com/datasets for dataset formatting guidance.
train: New cache created: /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train/train_images.cache WARNING ⚠️ No labels found in /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train/train_images.cache, training may not work correctly. See https://docs.ultralytics.com/datasets for dataset formatting guidance. albumentations: Blur(p=0.01, blur_limit=(3, 7)), MedianBlur(p=0.01, blur_limit=(3, 7)), ToGray(p=0.01), CLAHE(p=0.01, clip_limit=(1, 4.0), tile_grid_size=(8, 8)) val: Scanning /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train/train_images... 0 images, 86 backgrounds, 0 corrupt: 100%|██████████| 86/86 [00:40<00:00, 2.14it/s]val: WARNING ⚠️ No labels found in /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train/train_images.cache. See https://docs.ultralytics.com/datasets for dataset formatting guidance. val: New cache created: /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train/train_images.cache WARNING ⚠️ No labels found in /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train/train_images.cache, training may not work correctly. See https://docs.ultralytics.com/datasets for dataset formatting guidance.
Plotting labels to runs/detect/train8/labels.jpg... zero-size array to reduction operation maximum which has no identity optimizer: 'optimizer=auto' found, ignoring 'lr0=0.01' and 'momentum=0.937' and determining best 'optimizer', 'lr0' and 'momentum' automatically... optimizer: AdamW(lr=0.000156, momentum=0.9) with parameter groups 57 weight(decay=0.0), 64 weight(decay=0.0005), 63 bias(decay=0.0) TensorBoard: model graph visualization added ✅ Image sizes 640 train, 640 val Using 0 dataloader workers Logging results to runs/detect/train8 Starting training for 100 epochs...
Epoch GPU_mem box_loss cls_loss dfl_loss Instances Size 1/100 0G 0 127.3 0 0 640: 17%|█▋ | 8/48 [04:40<22:42, 34.05s/it]
Glenn JocherUltralytics@Mahmoud-owais the warning indicates that no labels were found in the specified directory for the xView dataset. This could be due to incorrect dataset formatting or missing label files. Please ensure that your dataset is correctly formatted and that the label files are in the expected location. For detailed guidance on dataset formatting, refer to the Ultralytics dataset documentation. Additionally, updating to the latest version of the Ultralytics package might resolve any underlying issues. You can update using
pip install -U ultralytics.1Mahmoud-owais
thanks datasets/ └── xView/ ├── images/ │ ├── train/ │ └── val/ ├── labels/ │ ├── train/ │ └── val/ i have the dataset and xView_train.geojson how to by xView.yaml to extract the xView_train.geojson to this structure datasets/ └── xView/ ├── images/ │ ├── train/ │ └── val/ ├── labels/ │ ├── train/ │ └── val/
Mahmoud-owais
and i found this Error Transferred 319/355 items from pretrained weights TensorBoard: Start with 'tensorboard --logdir runs/detect/train28', view at http://localhost:6006/ Freezing layer 'model.22.dfl.conv.weight' train: Scanning /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train... 0 images, 760 backgrounds, 3 corrupt: 100%|██████████| 763/763 [06:10<00:00, 2.06it/s]train: WARNING ⚠️ /content/drive/MyDrive/ultralytics-main/datasets/xView/images/train/._100.tif: ignoring corrupt image/label: cannot identify image file '/content/drive/MyDrive/ultralytics-main/datasets/xView/images/train/._100.tif' train: WARNING ⚠️ /content/drive/MyDrive/ultralytics-main/datasets/xView/images/train/._102.tif: ignoring corrupt image/label: cannot identify image file '/content/drive/MyDrive/ultralytics-main/datasets/xView/images/train/._102.tif' train: WARNING ⚠️ /content/drive/MyDrive/ultralytics-main/datasets/xView/images/train/._109.tif: ignoring corrupt image/label: cannot identify image file '/content/drive/MyDrive/ultralytics-main/datasets/xView/images/train/._109.tif' train: WARNING ⚠️ No labels found in /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train.cache. See https://docs.ultralytics.com/datasets for dataset formatting guidance.
train: New cache created: /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train.cache WARNING ⚠️ No labels found in /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train.cache, training may not work correctly. See https://docs.ultralytics.com/datasets for dataset formatting guidance. albumentations: Blur(p=0.01, blur_limit=(3, 7)), MedianBlur(p=0.01, blur_limit=(3, 7)), ToGray(p=0.01), CLAHE(p=0.01, clip_limit=(1, 4.0), tile_grid_size=(8, 8)) val: Scanning /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train... 0 images, 86 backgrounds, 0 corrupt: 100%|██████████| 86/86 [00:41<00:00, 2.07it/s]val: WARNING ⚠️ No labels found in /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train.cache. See https://docs.ultralytics.com/datasets for dataset formatting guidance. val: New cache created: /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train.cache WARNING ⚠️ No labels found in /content/drive/MyDrive/ultralytics-main/datasets/xView/labels/train.cache, training may not work correctly. See https://docs.ultralytics.com/datasets for dataset formatting guidance.
Plotting labels to runs/detect/train28/labels.jpg... zero-size array to reduction operation maximum which has no identity optimizer: 'optimizer=auto' found, ignoring 'lr0=0.01' and 'momentum=0.937' and determining best 'optimizer', 'lr0' and 'momentum' automatically... optimizer: AdamW(lr=0.000156, momentum=0.9) with parameter groups 57 weight(decay=0.0), 64 weight(decay=0.0005), 63 bias(decay=0.0) TensorBoard: model graph visualization added ✅ Image sizes 640 train, 640 val Using 0 dataloader workers Logging results to runs/detect/train28 Starting training for 100 epochs...
Epoch GPU_mem box_loss cls_loss dfl_loss Instances Size
Glenn JocherUltralytics@Mahmoud-owais it looks like the issue is related to corrupt image files and missing labels in your dataset. The warnings indicate that some images in your training set are not being recognized, and no labels were found in the specified directory. Please ensure that your dataset is correctly formatted and that all necessary label files are present. You can refer to the Ultralytics dataset documentation for detailed guidance on proper dataset formatting. Additionally, make sure to check if the issue persists with the latest version of the Ultralytics package.
1
goodbbboy
hello,I want to split original xView images with high-resolution into images with 800x800 resolution in multiscale way as DOTA, how should I modify "ultralytics/data/split_dota.py"? Now I have the .txt labels of xView by default YOLOv8 code, but their format is different from DOTA's txt file. Thanks a lot!!!
Glenn JocherUltralytics(bearbeitet)@goodbbboy hello!
To split the original xView images into 800x800 resolution tiles similar to the DOTA dataset, you'll need to modify the
ultralytics/data/split_dota.pyscript to handle the xView dataset format. Here are some steps to guide you through the process:-
Modify the Splitting Logic: Adjust the script to read the xView images and split them into 800x800 tiles. Ensure that the splitting logic handles the overlap and edge cases properly.
-
Adjust Label Conversion: Since the xView labels are in a different format, you'll need to convert them to match the DOTA format. This involves parsing the xView
.txtlabels and transforming them into the required format.
Here's a basic example to get you started:
import os from PIL import Image import numpy as np def split_image(image_path, output_dir, tile_size=800, overlap=200): image = Image.open(image_path) image_name = os.path.basename(image_path).split('.')[0] width, height = image.size for i in range(0, width, tile_size - overlap): for j in range(0, height, tile_size - overlap): box = (i, j, i + tile_size, j + tile_size) tile = image.crop(box) tile.save(os.path.join(output_dir, f"{image_name}_{i}_{j}.png")) def convert_labels(xview_label_path, output_dir, tile_size=800, overlap=200): with open(xview_label_path, 'r') as file: lines = file.readlines() for line in lines: # Parse the xView label format and convert to DOTA format # This is a placeholder; you'll need to implement the actual conversion logic pass # Example usage split_image('path/to/xview/image.tif', 'output/tiles') convert_labels('path/to/xview/label.txt', 'output/labels')- Integrate with Ultralytics YOLO: Ensure that the modified script integrates well with the Ultralytics YOLO training pipeline. You might need to adjust the dataset YAML file to point to the new split images and labels.
For more detailed guidance on creating a minimum reproducible example, please refer to our Minimum Reproducible Example documentation. Additionally, make sure you are using the latest version of the Ultralytics packages to benefit from the latest features and bug fixes.
Feel free to reach out if you have any further questions. Happy coding! 🚀
-
vonzo
Hello, the dataset is not being downloaded automaticaly when
yolo detect train data=xView.yaml model=yolov8n.pt epochs=100 imgsz=640I get
set 'xView.yaml' images not found ⚠️, missing path '/home/ltu-builder/GIT/datasets/xView/images/autosplit_val.txt' Traceback (most recent call last): File "/home/ltu-builder/miniconda3/lib/python3.11/site-packages/ultralytics/engine/trainer.py", line 514, in get_dataset data = check_det_dataset(self.args.data) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/ltu-builder/miniconda3/lib/python3.11/site-packages/ultralytics/data/utils.py", line 338, in check_det_dataset exec(s, {"yaml": data}) File "<string>", line 63, in <module> File "<string>", line 16, in convert_labels FileNotFoundError: [Errno 2] No such file or directory: '/home/ltu-builder/GIT/datasets/xView/xView_train.geojson' The above exception was the direct cause of the following exception: Traceback (most recent call last): File "/home/ltu-builder/miniconda3/bin/yolo", line 8, in <module> sys.exit(entrypoint()) ^^^^^^^^^^^^ File "/home/ltu-builder/miniconda3/lib/python3.11/site-packages/ultralytics/cfg/__init__.py", line 582, in entrypoint getattr(model, mode)(**overrides) # default args from model ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/ltu-builder/miniconda3/lib/python3.11/site-packages/ultralytics/engine/model.py", line 649, in train self.trainer = (trainer or self._smart_load("trainer"))(overrides=args, _callbacks=self.callbacks) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/ltu-builder/miniconda3/lib/python3.11/site-packages/ultralytics/engine/trainer.py", line 129, in __init__ self.trainset, self.testset = self.get_dataset() ^^^^^^^^^^^^^^^^^^ File "/home/ltu-builder/miniconda3/lib/python3.11/site-packages/ultralytics/engine/trainer.py", line 518, in get_dataset raise RuntimeError(emojis(f"Dataset '{clean_url(self.args.data)}' error ❌ {e}")) from e RuntimeError: Dataset 'xView.yaml' error ❌ [Errno 2] No such file or directory: '/home/ltu-builder/GIT/datasets/xView/xView_train.geojson'Can I force dowload? or do it by hand?
Paula DerrengerUltralyticsHello! 👋 It looks like you're encountering an issue with the automatic download of the xView dataset. The xView dataset requires manual downloading due to its size and licensing.
For this dataset, you need to:
- Visit the xView dataset website and download the dataset files manually.
- Follow the instructions for downloading
train_images.zipandval_images.zip. - Unzip these files under
'datasets/xView'relative to your working directory.
After downloading and organizing the dataset, your file structure should look something like this:
ultralytics/ └── datasets/ └── xView/ ├── train_images/ └── val_images/Make sure to place
xView.yamlcorrectly in your project, specifying this file structure.If you have further issues or questions, feel free to ask. Happy training! 🚀



