Skip to content

Latest commit

 

History

History
39 lines (27 loc) · 6.3 KB

README.md

File metadata and controls

39 lines (27 loc) · 6.3 KB
Raw

PSC

Phase-Shifting Coder: Predicting Accurate Orientation in Oriented Object Detection

Abstract

With the vigorous development of computer vision, oriented object detection has gradually been featured. In this paper, a novel differentiable angle coder named phase-shifting coder (PSC) is proposed to accurately predict the orientation of objects, along with a dual-frequency version PSCD. By mapping rotational periodicity of different cycles into phase of different frequencies, we provide a unified framework for various periodic fuzzy problems in oriented object detection. Upon such framework, common problems in oriented object detection such as boundary discontinuity and square-like problems are elegantly solved in a unified form. Visual analysis and experiments on three datasets prove the effectiveness and the potentiality of our approach. When facing scenarios requiring high-quality bounding boxes, the proposed methods are expected to give a competitive performance.

Results and models

DOTA1.0

Backbone AP50 AP75 Angle Dual-freqency lr schd Mem (GB) Inf Time (fps) Aug Batch Size Configs Download
ResNet50 (1024,1024,200) 71.14 40.28 le90 True 1x 4.29 21.9 - 2 rotated-fcos-hbox-le90_r50_fpn_psc-dual_1x_dota model | log
ResNet50 (1024,1024,200) 71.92 45.84 le90 True 1x 2.26 20.1 - 2 rotated-retinanet-rbox-le90_r50_fpn_psc-dual_amp-1x_dota (note1) model | log

note1: Trained with --cfg-options load_from=rotated_retinanet_obb_r50_fpn_1x_dota_le90-c0097bc4.pth

HRSC

Backbone AP50 AP75 Angle Dual-freqency lr schd Mem (GB) Inf Time (fps) Aug Batch Size Configs Download
ResNet50 (800,512) 90.10 79.39 le90 - 6x 1.46 37.0 RR 2 rotated-fcos-hbox-le90_r50_fpn_psc_rr-6x_hrsc model | log
ResNet50 (800,512) 84.86 60.07 le90 - 6x 1.44 35.7 RR 2 rotated-retinanet-rbox-le90_r50_fpn_psc_rr-6x_hrsc model | log

Citation

@inproceedings{yu2023psc,
    author = {Yu, Yi and Da, Feipeng},
    title = {Phase-Shifting Coder: Predicting Accurate Orientation in Oriented Object Detection},
    booktitle = {Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    year = {2023},
    url = {https://arxiv.org/abs/2211.06368}
}