AI Paper Atlas

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

Convolutions Die Hard: Open-Vocabulary Segmentation with Single Frozen Convolutional CLIP

Qihang Yu, Ju He, Xueqing Deng, Xiaohui Shen, Liang-Chieh Chen

NeurIPS 2023 · Detection and segmentation › Open-set and few-shot detection · approach: Open-vocabulary semantic segmentation

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
Use frozen CLIP for all three modules instead of trainable CLIP (ablation)FC-CLIP · ADE20K zero-shot panopticPQ: 24.1 -> 26.8Improved
Add frozen out-of-vocabulary classifier to trainable mask generator and in-vocab classifier (ablation)FC-CLIP · ADE20K zero-shot panopticPQ: 17.7 -> 25.4Improved
Trainable in-vocab classifier versus frozen out-of-vocab classifier on unseen classes (ablation)FC-CLIP · ADE20KPQ unseen: 12.6 -> 2.6Hurt
Share single frozen backbone instead of trainable plus frozen CLIP copies (ablation)FC-CLIP · ADE20K zero-shot panopticPQ: 25.4 -> 26.8Mixed
Geometric instead of arithmetic ensemble (ablation)FC-CLIP · ADE20K zero-shot panopticPQ: 25.6 -> 26.8Improved
Use alpha=0.4, beta=0.8 instead of in-vocab only (0.0, 0.0) (ablation)FC-CLIP · ADE20K zero-shot panopticPQ: 17.8 -> 26.8Improved
Use out-of-vocab only (alpha=1, beta=1) instead of tuned ensemble (ablation)FC-CLIP · ADE20K zero-shot panopticPQ: 26.8 -> 21.9Hurt
Swap seen/unseen weights (alpha=1.0, beta=0.0) (ablation)FC-CLIP · ADE20K zero-shot panopticPQ: 26.8 -> 17.5Hurt
Increase CLIP input resolution from 224 to 1120 with ConvNeXt-L (ablation)CLIP classifier · COCOPQ: 17.3 -> 29.3Improved
Increase CLIP input resolution from 224 to 1120 with ViT-L/14 (ablation)CLIP classifier · COCOPQ: 19.3 -> 14.9Hurt
Increase ImageNet input resolution from 224 to 896 with ViT-L/14 (ablation)CLIP classifier · ImageNet-1kaccuracy: 75.3 -> 53.9Hurt
Increase ImageNet input resolution from 224 to 336 with ConvNeXt-L (ablation)CLIP classifier · ImageNet-1kaccuracy: 75.1 -> 77.1Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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