mmdet-rgbtdroneperson

Official code for "Drone-based RGBT Tiny Person Detection".

Installation

Please refer to https://github.com/open-mmlab/mmdetection/tree/2.x

Environment

mmdet 2.25.1
mmcv-full 1.6.1
pytorch 1.10.0

Trained Model

On RGBTDronePerson

Models	mAP50	mAP50(tiny)	mAP25
QFDet	42.08	44.04	57.34
QFDet*	46.72	48.75	61.62

On VTUAV-det

Models	mAP	mAP50	mAP75
QFDet	31.10	70.40	22.90
QFDet*	33.30	75.50	24.20

Train

Train QFDet on RGBTDronePerson.

python tools/train.py qfdet_configs/qfdet_r50_fpn_1x_rgbtdroneperson.py

Train QFDet* on RGBTDronePerson.

python tools/train.py qfdet_configs/qfdet_star_r50_fpn_1x_rgbtdroneperson.py

Train QFDet on VTUAV-det.

python tools/train.py qfdet_configs/qfdet_r50_fpn_1x_rgbtdroneperson.py

Train QFDet* on VTUAV-det.

python tools/train.py qfdet_configs/qfdet_star_r50_fpn_1x_rgbtdroneperson.py

Test

For example, test checkpoint epoch_11_qfdet_rgbtdroneperson.pth:

python tools/test.py qfdet_configs/qfdet_r50_fpn_1x_rgbtdroneperson.py work_dir/qfdet_r50_fpn/rgbtdroneperson/epoch_11_qfdet_rgbtdroneperson.pth --eval bbox

Dataset

Please refer to our github page.

Citation

@article{ZHANG202361,
title = {Drone-based RGBT tiny person detection},
journal = {ISPRS Journal of Photogrammetry and Remote Sensing},
volume = {204},
pages = {61-76},
year = {2023},
doi = {https://doi.org/10.1016/j.isprsjprs.2023.08.016},
url = {https://www.sciencedirect.com/science/article/pii/S0924271623002319},
author = {Yan Zhang and Chang Xu and Wen Yang and Guangjun He and Huai Yu and Lei Yu and Gui-Song Xia}
}