AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

Open-Vocabulary SAM: Segment and Recognize Twenty-thousand Classes Interactively

Haobo Yuan, Xiangtai Li, Chong Zhou, Yining Li, Kai Chen, Chen Change Loy

ECCV 2024 · Detection and segmentation › Open-set and few-shot detection · approach: Open-vocabulary semantic segmentation · 90 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
share one frozen CLIP backbone via SAM2CLIP distillation instead of separate SAM and CLIP backbonesOpen-Vocabulary SAM · COCO open-vocab, GT box promptsFLOPs / parameters: 3,545G -> 1,180G; 808M -> 304MImproved
add CLIP2SAM on top of SAM2CLIP (ablation)Open-Vocabulary SAM · COCO with Detic box promptsAP: 28.7 -> 34.4Improved
use SAM2CLIP alone (shared backbone) without CLIP2SAM (ablation)Open-Vocabulary SAM · COCO with Detic box promptsAP: 29.5 -> 28.7Hurt
use light FPN with RoIAlign for CLIP2SAM instead of a classification token with cross attention (ablation)CLIP2SAM · COCO open-vocab, GT box promptsRecognition accuracy: 79.3 -> 84.3Improved
use multi-scale instead of single-scale features in the SAM2CLIP adapter (ablation)SAM2CLIP · COCO open-vocab, GT box promptsIoU: 73.6 -> 78.3Improved
use ViT-based CLIP as frozen backbone instead of convolutional CLIP (ablation)Open-Vocabulary SAM · COCO open-vocab, GT box promptsIoU: 78.1 -> 38.6Hurt
use larger CLIP backbone RN50x64 instead of RN50x16 (ablation)Open-Vocabulary SAM · COCO open-vocab, GT box promptsRecognition accuracy: 55.1 -> 54.1No clear effect
add CoOp-style adapters to combined SAM and CLIP baselines (ablation)Feature-Crop baseline · COCO open-vocab, GT box promptsRecognition accuracy: 55.1 -> 70.9Improved
scale training data to V3Det + LVIS + I-21k (ablation)Open-Vocabulary SAM · Large-vocabulary scalingAccuracy: 83.1 -> 83.3No clear effect
add Object365 to V3Det + LVIS + I-21k training (ablation)Open-Vocabulary SAM · Large-vocabulary scalingAccuracy: 83.3 -> 83.0No clear effect
Open-Vocabulary SAM vs SAM-Huge on LVIS mask quality with DeticOpen-Vocabulary SAM · LVIS mask quality with Deticmask mAP: 43.5 -> 42.6Hurt

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