The implementation of Gradient Harmonized Single-stage Detector published on AAAI 2019 (Oral).
- The GHM-C and GHM-R loss functions are available in ghm_loss.py.
- The code works for pytorch 0.4.1 and later version. If you want to run it with pytorch 0.3.x, please checkout to the pytorch-0.3 branch.
- The main training code is based on mmdetection. Please see this for installation issues.
- We provide training and testing scripts and configuration files for both GHM and baseline (focal loss and smooth L1 loss) in the experiments directory.
The use of this code is RESTRICTED to non-commercial research and educational purposes.
@article{li2019ghm,
title={Gradient Harmonized Single-stage Detector},
author={Buyu Li, Yu Liu, Xiaogang Wang},
booktitle={AAAI},
year={2019}
}