SASM icon indicating copy to clipboard operation
SASM copied to clipboard

AAAI22-2171-SASM

Source Code of AAAI22-2171

Introduction

The source code includes training and inference procedures for the proposed method of the paper submitted to the Thirty-Sixth AAAI Conference on Artificial Intelligence (AAAI 2022) with title "Shape-Adaptive Selection and Measurement for Oriented Object Detection" (ID: 2171).

The the effectiveness of the proposed method is verified on two baseline methods. Corresponding source code and configurations reside in following two sub-directories:

  • sasm_reppoints: the implementation and verification using RepPoints as baseline;
  • sasm_s2anet: the implementation and verification using s2anet(https://ieeexplore.ieee.org/document/9377550) as baseline.

We provide only the source code related to the proposed method in the sub-directories so that reviewers can check them quickly and conveniently.

Please refer to the README.md file in each sub-directory for the detailed instructions of usage.

Introduction

Method Assignment Reg. Loss Tricks mAP
RepPoints MaxIoU GIoU - 70.46
RepPoints SASM BCLoss + GIoU - 74.27
RepPoints SASM BCLoss + GIoU MS training 77.19
s2anet SASM Smooth L1 MS training 79.17

Reference

1、https://github.com/open-mmlab/mmdetection

2、https://github.com/LiWentomng/OrientedRepPoints

3、https://github.com/SDL-GuoZonghao/BeyondBoundingBox

4、https://github.com/csuhan/s2anet

5、https://github.com/sfzhang15/ATSS