AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

CASTing Your Model: Learning To Localize Improves Self-Supervised Representations

Ramprasaath R. Selvaraju, Karan Desai, Justin Johnson, Nikhil Naik

CVPR 2021 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Data-efficient and robust self-supervision · 59 citations (OpenAlex)

Check 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 saliency-constrained crops and Grad-CAM attention lossMoCo pretrained on COCO · VOC detectionAPbbox all: 47.5 -> 54.2Improved
Add saliency-constrained crops and Grad-CAM attention lossMoCo pretrained on COCO · VOC classificationmAP: 66.9 -> 73.1Improved
Add saliency-constrained crops and Grad-CAM attention lossMoCo pretrained on COCO · ImageNet-1k classificationTop-1 accuracy: 46.5 -> 48.7Improved
Add saliency-constrained crops and Grad-CAM attention lossMoCo pretrained on COCO · COCO valMean localization IoU: 0.24 -> 0.41Improved
Add saliency-constrained crops without attention loss (ablation)MoCo pretrained on COCO · VOC classificationmAP: 66.9 -> 69.3Improved
Add saliency-constrained crops without attention loss (ablation)MoCo pretrained on COCO · ImageNet-1k classificationTop-1 accuracy: 46.5 -> 46.0Hurt
Replace saliency-constrained cropping with random crops (phi=0.0) (ablation)MoCo-COCO + CAST · VOC classificationmAP: 73.1 -> 72.3Hurt
Replace saliency-constrained cropping with random crops (phi=0.0) (ablation)MoCo-COCO + CAST · ImageNet-1k classificationTop-1 accuracy: 48.7 -> 47.4Hurt
Add attention loss with lambda=3.0 instead of lambda=0.0 (ablation)MoCo-COCO with constrained crops · VOC classificationmAP: 66.9 -> 73.1Improved
Increase attention-loss weight to lambda=5.0 from lambda=3.0 (ablation)MoCo-COCO + CAST · VOC classificationmAP: 73.1 -> 72.4Hurt
Replace linear projection with a 2-layer MLP (ablation)MoCo-COCO + CAST · ImageNet-1k classificationTop-1 accuracy: 48.7 -> 50.1Improved
Restrict attention supervision to the query-key crop intersection (ablation)MoCo-COCO + CAST · VOC classificationmAP: 73.1 -> 72.0Hurt

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