v7.0.3

🤗 HuggingFace Hub Integration

• Use yolov5 models from hub:

import yolov5

# load model
model = yolov5.load('fcakyon/yolov5s-v7.0')

# set image
img = 'https://github.com/ultralytics/yolov5/raw/master/data/images/zidane.jpg'

# perform inference
results = model(img, size=640)

# show detection bounding boxes on image
results.show()

• Fine-tune yolov5 models from hub:

yolov5 train --img 640 --batch 16 --weights fcakyon/yolov5s-v7.0 --epochs 10 --device cuda:0

• Automatically push fine-tuned weight and training logs to hub (with autogenerated model card):

yolov5 train --data data.yaml --weights yolov5s.pt --hf_model_id username/modelname --hf_token YOUR-HF-WRITE-TOKEN

Available models: https://huggingface.co/models?other=yolov5

What's Changed

• ad hf hub tests to package testing by @fcakyon in #177
• fix private hub model download by @fcakyon in #179
• add automatic hf hub upload, refactor helpers by @fcakyon in #180
• fix readme by @fcakyon in #181

Full Changelog: 7.0.2...7.0.3

v7.0.2

What's Changed

• fix coco dataset training support by @fcakyon in #174

Full Changelog: 7.0.1...7.0.2

v7.0.1

What's Changed

• support loading models from huggingface hub by @fcakyon in #172

Full Changelog: 7.0.0...7.0.1

v7.0.0

What's Changed

• better exception handling for hublike loading by @fcakyon in #167
• update to ultralytics/yolov5 13.12.22 by @fcakyon in #170

Full Changelog: 6.2.3...7.0.0

v6.2.3

What's Changed

• update to ultralytics/yolov5 26.10.22 by @fcakyon in #164

Full Changelog: 6.2.2...6.2.3

v6.2.2

What's Changed

• update workflow versions by @fcakyon in #156
• fix a bug in train resume by @QazyBi in #161

New Contributors

• @QazyBi made their first contribution in #161

Full Changelog: 6.2.1...6.2.2

v6.2.1

What's Changed

• fix package_testing by @fcakyon in #148
• Fixed Invalid CUDA error. by @kadirnar in #149
• revert dataset path resolve by @fcakyon in #151
• fix segment dataloading error by @fcakyon in #153
• update version by @fcakyon in #155

New Contributors

• @kadirnar made their first contribution in #149

Full Changelog: 6.2.0...6.2.1

v6.2.0

New Features

Classify

Train/Val/Predict with YOLOv5 image classifier:

$ yolov5 classify train --img 640 --data mnist2560 --weights yolov5s-cls.pt --epochs 1

$ yolov5 classify predict --img 640 --weights yolov5s-cls.pt --source images/

from yolov5.classify import train, val, predict

train.run(imgsz=640, data='mnist2560.yaml')
val.run(imgsz=640, weights='yolov5s-cls.pt')
predict.run(imgsz=640)

import yolov5

model = yolov5.load('yolov5s-cls.pt')

Segment

Train/Val/Predict with YOLOv5 instance segmentation model:

$ yolov5 segment train --img 640 --weights yolov5s-seg.pt --epochs 1

$ yolov5 segment predict --img 640 --weights yolov5s-seg.pt --source images/

from yolov5.segment import train, val, predict

train.run(imgsz=640, data='coco128.yaml')
val.run(imgsz=640, weights='yolov5s-seg.pt')
predict.run(imgsz=640)

import yolov5

model = yolov5.load('yolov5s-seg.pt')

What's Changed

• Fix pypi package version not updating in readme by @Isydmr in #141
• fix typo notebook utils by @fcakyon in #144
• update export arg usage by @fcakyon in #145
• update to ultralytics 16.09.22 by @fcakyon in #146
• add classify and segment usage into readme by @fcakyon in #147

New Contributors

• @Isydmr made their first contribution in #141

Full Changelog: 6.1.9...6.2.0

v6.1.9

What's Changed

• fix evolve by @fcakyon in #139

Full Changelog: 6.1.8...6.1.9

v6.1.8

What's Changed

• remove redundant modules by @fcakyon in #133

Full Changelog: 6.1.7...6.1.8