TridentNet in Detectron2

Scale-Aware Trident Networks for Object Detection

Yanghao Li*, Yuntao Chen*, Naiyan Wang, Zhaoxiang Zhang

[TridentNet] [arXiv] [BibTeX]

In this repository, we implement TridentNet-Fast in Detectron2. Trident Network (TridentNet) aims to generate scale-specific feature maps with a uniform representational power. We construct a parallel multi-branch architecture in which each branch shares the same transformation parameters but with different receptive fields. TridentNet-Fast is a fast approximation version of TridentNet that could achieve significant improvements without any additional parameters and computational cost.

Training

To train a model, run

python /path/to/detectron2/projects/TridentNet/train_net.py --config-file <config.yaml>

For example, to launch end-to-end TridentNet training with ResNet-50 backbone on 8 GPUs, one should execute:

python /path/to/detectron2/projects/TridentNet/train_net.py --config-file configs/tridentnet_fast_R_50_C4_1x.yaml --num_gpus 8

Evaluation

Model evaluationcan be done similarly:

python /path/to/detectron2/projects/TridentNet/train_net.py --config-file configs/tridentnet_fast_R_50_C4_1x.yaml --eval-only MODEL.WEIGHTS model.pth

Results on MS-COCO in Detectron2

Model	Backbone	Head	lr sched	AP	AP50	AP75	APs	APm	APl
Faster	R50-C4	C5-512ROI	1X	35.7	56.1	38.0	19.2	40.9	48.7
TridentFast	R50-C4	C5-128ROI	1X	37.9	57.8	40.7	19.7	42.1	54.2
Faster	R50-C4	C5-512ROI	3X	38.4	58.7	41.3	20.7	42.7	53.1
TridentFast	R50-C4	C5-128ROI	3X	41.0	60.9	44.2	22.7	45.2	57.0
Faster	R101-C4	C5-512ROI	3X	41.1	61.4	44.0	22.2	45.5	55.9
TridentFast	R101-C4	C5-128ROI	3X	43.4	62.9	46.6	24.2	47.9	59.9

Citing TridentNet

If you use TridentNet, please use the following BibTeX entry.

@InProceedings{li2019scale,
  title={Scale-Aware Trident Networks for Object Detection},
  author={Li, Yanghao and Chen, Yuntao and Wang, Naiyan and Zhang, Zhaoxiang},
  journal={The International Conference on Computer Vision (ICCV)},
  year={2019}
}