Skip to content

Latest commit

 

History

History

tridentnet

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
README.md
README.md
 
 
__init__.py
__init__.py
 
 
builder.py
builder.py
 
 
builder_v2.py
builder_v2.py
 
 
input.py
input.py
 
 
resnet_v1.py
resnet_v1.py
 
 
resnet_v1b.py
resnet_v1b.py
 
 
resnet_v2.py
resnet_v2.py
 
 

README.md

Scale-Aware Trident Networks for Object Detection

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

Introduction

This repository implements TridentNet in the SimpleDet framework.

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. Then, we propose a scale-aware training scheme to specialize each branch by sampling object instances of proper scales for training. As a bonus, a fast approximation version of TridentNet could achieve significant improvements without any additional parameters and computational cost. On the COCO dataset, our TridentNet with ResNet-101 backbone achieves state-of-the-art single-model results by obtaining an mAP of 48.4.

Trident Blocks

  • Dilated convolution for efficient scale enumeration
  • Weight sharing between convs for uniform representation

The above figure shows how to convert bottleneck residual blocks to 3-branch Trident Blocks. The dilation rate of three branches are set as 1, 2 and 3, respectively.

Use TridentNet

Please setup SimpleDet following README and INSTALL and use the TridentNet configuration files in the config folder.

Results on MS-COCO

Backbone Test data mAP@[0.5:0.95] Link
Faster R-CNN, 1x ResNet-101 minival 37.6 model
TridentNet, 1x ResNet-101 minival 40.6 model
TridentNet, 1x, Fast Approx ResNet-101 minival 39.9 model
TridentNet, 2x ResNet-101 test-dev 42.8 model
TridentNet*, 3x ResNet-101 test-dev 48.4 model

Note:

  1. These models are not trained in SimpleDet. Re-training these models in SimpleDet gives a slightly better result.
  2. TridentNet* - TridentNet = extended training + softNMS + multi-scale training/testing + syncBN + DCNv1.

Results on MS-COCO with stronger baselines

All config files are available in config/resnet_v1b.

Model Backbone Head Train Schedule AP AP50 AP75 APs APm APl
Faster R50v1b-C4 C5-512ROI 2X 36.9 57.9 39.3 19.9 41.4 50.2
Trident R50v1b-C4 C5-128ROI 2X 39.6 60.9 42.9 22.5 44.5 53.9
TridentFast R50v1b-C4 C5-128ROI 2X 39.0 60.2 41.8 20.8 43.6 53.8
Faster R101v1b-C4 C5-512ROI 2X 40.5 61.2 43.8 22.5 44.8 55.4
Trident R101v1b-C4 C5-128ROI 2X 43.0 64.3 46.3 25.3 47.9 58.4
TridentFast R101v1b-C4 C5-128ROI 2X 42.5 63.7 46.0 23.3 46.7 59.3
Faster R152v1b-C4 C5-512ROI 2X 41.8 62.4 45.2 23.2 46.0 56.9
Trident R152v1b-C4 C5-128ROI 2X 44.4 65.4 48.3 26.4 49.4 59.6
TridentFast R152v1b-C4 C5-128ROI 2X 43.9 65.1 47.0 25.1 48.1 60.4

Citing TridentNet

@article{li2019scale,
  title={Scale-Aware Trident Networks for Object Detection},
  author={Li, Yanghao and Chen, Yuntao and Wang, Naiyan and Zhang, Zhaoxiang},
  journal={ICCV 2019},
  year={2019}
}