CoupleNet: Coupling Global Structure with Local Parts for Object Detection
The Code is modified from py-R-FCN, please follow the procedure in it to prepare the training data and testing data.
Using the default hyperparameters and iterations, you can achieve a mAP around 81.7%.
| training data | test data | [email protected] | time/img(ms) | |
|---|---|---|---|---|
| CoupleNet, ResNet-101** | VOC 07+12 | VOC 07 test | 81.7% | 102 |
| CoupleNet, ResNet-101 | VOC 07+12 | VOC 07 test | 82.1% | 122 |
| CoupleNet, ResNet-101 | VOC 07++12 | VOC 12 test | 80.4% | 122 |
**: without adding context.
| training data | test data | mAP@[0.5:0.95] | time/img(ms) | |
|---|---|---|---|---|
| CoupleNet, ResNet-101 | COCO 2014 trainval | COCO test dev | 34.4% | 122 |
VOC 0712 model (trained on VOC 07+12, mAP 81.7%)
If you find CoupleNet useful in your research, please consider citing:
@article{zhu2017couplenet,
title={CoupleNet: Coupling Global Structure with Local Parts for Object Detection},
author={Zhu, Yousong and Zhao, Chaoyang and Wang, Jinqiao and Zhao, Xu and Wu, Yi and Lu, Hanqing},
journal={arXiv preprint arXiv:1708.02863},
year={2017}
}