Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

v2.0 Release #491

Merged
merged 1 commit into from
Jul 23, 2020
Merged

v2.0 Release #491

merged 1 commit into from
Jul 23, 2020

Conversation

glenn-jocher
Copy link
Member

@glenn-jocher glenn-jocher commented Jul 23, 2020

Signed-off-by: Glenn Jocher glenn.jocher@ultralytics.com

πŸ› οΈ PR Summary

Made with ❀️ by Ultralytics Actions

🌟 Summary

YOLOv5 major updates with enhanced model definitions and performance improvements.

πŸ“Š Key Changes

  • Updated YOLOv5 pretrained checkpoints with slightly different performance metrics (AP values and model names).
  • Modified Detect layer initialization to accommodate new channel count parameter (ch).
  • Refined anchor order check to ensure consistency with stride order.
  • Adjusted values in training hyperparameters, such as class loss gain and HSV augmentation strengths.
  • Implemented a cosmetic change to README.md on model public release and development start dates.
  • Modified LR scheduler function for a new learning rate cosine annealing strategy.
  • Increased balancing factors in loss computation to tweak the influence of each prediction head.
  • Changed batch normalization momentum and epsilon values in utils/torch_utils.py.

🎯 Purpose & Impact

  • πŸš€ Enhanced performance: Tweaks in model and loss configuration can lead to more accurate detection and training stability.
  • πŸ€– Model training: Hyperparameter adjustments and learning rate scheduler optimization for better convergence.
  • πŸ‘€ Clearer documentation: README update helps users to follow the timeline of significant changes and releases.
  • 🧠 Backend improvements: Code modifications and cleanup contribute to maintenance and potential performance gains.

Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>
@glenn-jocher
Copy link
Member Author

This v2.0 release PR is complete, all CI CPU tests are passing. Will merge and subsequently issue v2.0 release.

IMPORTANT: v2.0 release contains breaking changes. Models trained with earlier versions will not operate correctly with v2.0. The last commit before v2.0 that operates correctly with all earlier pretrained models is:
https://github.com/ultralytics/yolov5/tree/5e970d45c44fff11d1eb29bfc21bed9553abf986

@glenn-jocher
Copy link
Member Author

Commands to clone last commit prior to v2.0:

git clone https://github.com/ultralytics/yolov5  # clone repo
cd yolov5
git reset --hard 5e970d4  # last commit before v2.0

@glenn-jocher glenn-jocher merged commit 9da56b6 into master Jul 23, 2020
@glenn-jocher glenn-jocher deleted the v2 branch July 23, 2020 22:46
BjarneKuehl pushed a commit to fhkiel-mlaip/yolov5 that referenced this pull request Aug 26, 2022
Signed-off-by: Glenn Jocher <glenn.jocher@ultralytics.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

1 participant