Skip to content

Conversation

glenn-jocher
Copy link
Member

@glenn-jocher glenn-jocher commented Apr 3, 2022

Eliminate 1 op during training and inference.

πŸ› οΈ PR Summary

Made with ❀️ by Ultralytics Actions

🌟 Summary

Improved bounding box predictions in YOLOv5 πŸ”.

πŸ“Š Key Changes

  • Corrected the equation used to calculate the x and y coordinates of bounding boxes.
  • Introduced a grid offset of -0.5 when calculating grid values for positioning.

🎯 Purpose & Impact

  • 🎯 Purpose: These modifications ensure more accurate and reliable bounding box outputs from the model by refining the coordinate calculations.
  • 🌐 Impact: Users of the YOLOv5 model will experience better object detection performance with potentially higher precision in localizing objects.

Eliminate 1 op during training and inference.
@glenn-jocher glenn-jocher self-assigned this Apr 3, 2022
@glenn-jocher glenn-jocher merged commit ea72b84 into master Apr 3, 2022
@glenn-jocher glenn-jocher deleted the update/grid branch April 3, 2022 21:40
BjarneKuehl pushed a commit to fhkiel-mlaip/yolov5 that referenced this pull request Aug 26, 2022
Eliminate 1 op during training and inference.
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.

1 participant