English

TDAF: Top-Down Attention Framework for Vision Tasks

Computer Vision and Pattern Recognition 2020-12-15 v1

Abstract

Human attention mechanisms often work in a top-down manner, yet it is not well explored in vision research. Here, we propose the Top-Down Attention Framework (TDAF) to capture top-down attentions, which can be easily adopted in most existing models. The designed Recursive Dual-Directional Nested Structure in it forms two sets of orthogonal paths, recursive and structural ones, where bottom-up spatial features and top-down attention features are extracted respectively. Such spatial and attention features are nested deeply, therefore, the proposed framework works in a mixed top-down and bottom-up manner. Empirical evidence shows that our TDAF can capture effective stratified attention information and boost performance. ResNet with TDAF achieves 2.0% improvements on ImageNet. For object detection, the performance is improved by 2.7% AP over FCOS. For pose estimation, TDAF improves the baseline by 1.6%. And for action recognition, the 3D-ResNet adopting TDAF achieves improvements of 1.7% accuracy.

Keywords

Cite

@article{arxiv.2012.07248,
  title  = {TDAF: Top-Down Attention Framework for Vision Tasks},
  author = {Bo Pang and Yizhuo Li and Jiefeng Li and Muchen Li and Hanwen Cao and Cewu Lu},
  journal= {arXiv preprint arXiv:2012.07248},
  year   = {2020}
}

Comments

Conference paper in AAAI 2021

R2 v1 2026-06-23T20:56:26.499Z