AI Paper Atlas

Research map / Detection and segmentation / CNN features

TDAF: Top-Down Attention Framework for Vision Tasks

Bo Pang, Yizhuo Li, Jiefeng Li, Muchen Li, Hanwen Cao, Cewu Lu

AAAI 2021 · Detection and segmentation › CNN features · approach: Convolution-transformer backbones · 11 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add top-down ANAR attention to VGG16VGG16 · CIFAR-10 testaccuracy: 92.64 -> 93.43Improved
Replace ANAR-3's learnable DeCNN up-sampling with parameter-free interpolation (ANAR-2) (ablation)VGG16 · CIFAR-10 testaccuracy: 93.43 -> 93.02Hurt
Add top-down ANAR attentionImageNet ResNet50 · ImageNet-2012 validationTop-1 accuracy: 77.1 -> 78.4Improved
Increase ANAR depth from 3 to 7 layers (ablation)ImageNet ResNet50 with 3 flows · ImageNet-2012 validationTop-1 accuracy: 78.4 -> 78.8Improved
Increase recursive flows from 3 to 4 (ablation)ImageNet ResNet101 with ANAR-3 · ImageNet-2012 validationTop-1 accuracy: 79.9 -> 80.2Improved
Remove inter-flow attention and concatenate standalone flow features (ablation)ImageNet ResNet50, 3-flow R2DNS · ImageNet-2012 validationTop-1 accuracy: 78.6 -> 77.5Hurt
Add top-down attention to FCOSResNet50-FCOS · COCO-2017AP: 37.1 -> 39.8Improved
Increase recursive flows from 3 to 4 (ablation)COCO ResNet101-FCOS with ANAR-3 · COCO-2017AP: 42.0 -> 42.3Improved
Increase ANAR depth from 3 to 5 layers (ablation)COCO ResNet50-FCOS with 3 flows · COCO-2017AP: 39.6 -> 39.8Improved
Add top-down attention to SimplePose ResNet50SimplePose ResNet50 · COCO-2017 validation, pose (GT detector)AP: 72.7 -> 74.3Improved
Increase ANAR depth from 3 to 7 layers (ablation)SimplePose ResNet50 with 3 flows · COCO-2017 validation, pose (GT detector)AP: 73.8 -> 74.3Improved
Add TDAF attention3D-ResNet50, pre-trained · Kinetics-400 validationaccuracy: 68.9 -> 70.1Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy