ethanhe42/KL-Loss
A PyTorch implementation of KL-Loss for improved bounding box regression in object detection using learned localization uncertainty.
Not currently ranked — collecting fresh signals.
star history
This repository implements a CVPR 2019 paper that proposes a novel bounding box regression loss enabling simultaneous learning of bounding box transformation and localization variance. The approach leverages learned variance during non-maximum suppression to merge neighboring boxes, improving localization accuracy. The implementation supports Faster R-CNN with VGG-16 and ResNet-50-FPN Mask R-CNN architectures, achieving significant AP improvements on MS-COCO.
Frequently asked
- What is ethanhe42/KL-Loss?
- A PyTorch implementation of KL-Loss for improved bounding box regression in object detection using learned localization uncertainty.
- Is KL-Loss open source?
- Yes — ethanhe42/KL-Loss is open source, released under the Apache-2.0 license.
- What language is KL-Loss written in?
- ethanhe42/KL-Loss is primarily written in Python.
- How popular is KL-Loss?
- ethanhe42/KL-Loss has 720 stars on GitHub.
- Where can I find KL-Loss?
- ethanhe42/KL-Loss is on GitHub at https://github.com/ethanhe42/KL-Loss.