Detection-Utils

Common functionality for object detection: This repository hosts functions that are commonly used across object detection projects.

The functionality here includes

• Box overlap computation (IoU)
• Precision and recall calculations
• Computing targets for training a detector given a set of ground-truth objects
• Non-maximum suppression
• Coordinate transform utilities
• Focal loss in PyTorch

All the functions here are well-tested to ensure proper functionality of these utilities. This repository is meant to ensure that modifications and improvements that are implemented as part of one program migrate to other programs when appropriate.