Architecture | Softmax, RPN, Balanced L1 Loss, Balanced Feature Pyramid, Convolution, FPN, RoIPool, IoU-Balanced Sampling, ResNet |
---|---|
lr sched | 1x |
SHOW MORE |
Architecture | Softmax, RPN, Balanced L1 Loss, Balanced Feature Pyramid, Convolution, FPN, RoIPool, IoU-Balanced Sampling, ResNet |
---|---|
lr sched | 1x |
SHOW MORE |
Architecture | Softmax, RPN, ResNeXt, Balanced L1 Loss, Balanced Feature Pyramid, Convolution, FPN, RoIPool, IoU-Balanced Sampling |
---|---|
lr sched | 1x |
SHOW MORE |
Architecture | Balanced L1 Loss, Balanced Feature Pyramid, FPN, IoU-Balanced Sampling, ResNet, Focal Loss |
---|---|
lr sched | 1x |
SHOW MORE |
[ALGORITHM]
We provide config files to reproduce the results in the CVPR 2019 paper Libra R-CNN.
@inproceedings{pang2019libra,
title={Libra R-CNN: Towards Balanced Learning for Object Detection},
author={Pang, Jiangmiao and Chen, Kai and Shi, Jianping and Feng, Huajun and Ouyang, Wanli and Dahua Lin},
booktitle={IEEE Conference on Computer Vision and Pattern Recognition},
year={2019}
}
The results on COCO 2017val are shown in the below table. (results on test-dev are usually slightly higher than val)
Architecture | Backbone | Style | Lr schd | Mem (GB) | Inf time (fps) | box AP | Config | Download |
---|---|---|---|---|---|---|---|---|
Faster R-CNN | R-50-FPN | pytorch | 1x | 4.6 | 19.0 | 38.3 | config | model | log |
Fast R-CNN | R-50-FPN | pytorch | 1x | |||||
Faster R-CNN | R-101-FPN | pytorch | 1x | 6.5 | 14.4 | 40.1 | config | model | log |
Faster R-CNN | X-101-64x4d-FPN | pytorch | 1x | 10.8 | 8.5 | 42.7 | config | model | log |
RetinaNet | R-50-FPN | pytorch | 1x | 4.2 | 17.7 | 37.6 | config | model | log |
MODEL | BOX AP |
---|---|
Faster R-CNN Libra R-CNN (X-101-64x4d-FPN, 1x, pytorch) | 42.7 |
Faster R-CNN Libra R-CNN (R-101-FPN, 1x, pytorch) | 40.1 |
Faster R-CNN Libra R-CNN (R-50-FPN, 1x, pytorch) | 38.3 |
RetinaNet Libra R-CNN (R-50-FPN, 1x, pytorch) | 37.6 |