AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

Perceptual Grouping in Contrastive Vision-Language Models

Kanchana Ranasinghe, Brandon McKinzie, Sachin Ravi, Yinfei Yang, Alexander Toshev, Jonathon Shlens

ICCV 2023 · Vision-language and multimodal › CLIP and zero-shot · approach: Open-vocabulary alignment from VLMs · 63 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
replace class token with max pooling for spatial aggregation (ablation)CLIPpy · PASCAL VOC val, CC-12MmIoU: 4.0 -> 50.8Improved
replace average pooling with max pooling (ablation)CLIPpy · PASCAL VOC val, CC-12MmIoU: 11.6 -> 50.8Improved
replace class token or average pooling with max pooling (ablation)CLIPpy · ImageNet zero-shot, CC-12Mtop-1 accuracy: 46.0 -> 42.3Hurt
initialize image encoder with DINO instead of supervised ImageNet-1K (ablation)CLIPpy · PASCAL VOC val, CC-12MmIoU: 22.5 -> 50.8Improved
initialize image encoder with DINO instead of random (ablation)CLIPpy · PASCAL VOC val, CC-12MmIoU: 23.5 -> 50.8Improved
add Sentence-T5 text initialization (ablation)CLIPpy · PASCAL VOC val, CC-12MmIoU: 44.3 -> 50.8Improved
freeze the pretrained image backbone (ablation)CLIPpy · PASCAL VOC val, CC-12MmIoU: 50.8 -> 10.4Hurt
freeze the pretrained text backbone (ablation)CLIPpy · PASCAL VOC val, CC-12MmIoU: 50.8 -> 29.5Hurt
add visual token sub-sampling during training (ablation)CLIPpy · PASCAL VOC valmIoU: 50.9 -> 51.8Improved
apply CLIPpy modifications versus CLIP baselineCLIP ViT-B/16 on CC-12M · ImageNet zero-shottop-1 accuracy: 46.0 -> 45.3No clear effect
apply CLIPpy modifications versus CLIP baselineCLIP ViT-B/16 on CC-12M · ADE20K and COCO bottom-up groupingJaccard Similarity: 22.9 -> 28.9 (ADE20K), 20.4 -> 26.0 (COCO)Improved
use CLIPpy grouping rather than baseline CLIP for zero-shot WaterbirdsCLIP · Waterbirds test, waterbird accuracy on incommensurate backgroundaccuracy drop (domain gap): CLIP 80.2 -> 48.1; CLIPpy 76.9 -> 74.9Improved

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