AI Paper Atlas

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

DualCoOp++: Fast and Effective Adaptation to Multi-Label Recognition With Limited Annotations

Ping Hu, Ximeng Sun, Stan Sclaroff, Kate Saenko

TPAMI 2023 · Vision-language and multimodal › CLIP and zero-shot · approach: Open-vocabulary alignment from VLMs · 42 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 evidence-guided spatial aggregation (ablation)DualCoOp · MS-COCO partial labels 50% (ablation)overall F1: 78.5 -> 79.7Improved
add winner-take-all module (ablation)DualCoOp++ · MS-COCO partial labels 50% (ablation)overall F1: 79.7 -> 80.4Improved
add evidence-guided aggregation (ablation)DualCoOp · MS-COCO zero-shot ZSLF1: 50.3 -> 51.2Improved
add WTA module (ablation)DualCoOp++ w/o WTA · MS-COCO zero-shot ZSLF1: 51.2 -> 52.5Improved
add WTA module (ablation)DualCoOp++ w/o WTA · NUS-WIDE zero-shot ZSL (Top-3)F1: 41.7 -> 46.9Improved
add negative prompt to positive prompt (ablation)Learnable prompt · MS-COCO zero-shot ZSLF1: 45.4 -> 50.3Improved
add evidential prompt to positive and negative prompts (ablation)DualCoOp · MS-COCO zero-shot ZSLF1: 50.3 -> 51.2Improved
use learnable positive prompt instead of hand-crafted prompt (ablation)CLIP · MS-COCO zero-shot ZSLF1: 36.5 -> 45.4Improved
use learnable negative-only prompt instead of positive-only (ablation)CLIP · MS-COCO zero-shot ZSLF1: 45.4 -> 13.5Hurt
use evidence-guided aggregation instead of multi-headed attention pooling (ablation)CLIP RN101 · MS-COCO partial labels 10%mAP: 70.4 -> 76.2Improved
test with 448 input after training at 224 using multi-headed attention pooling (ablation)CLIP RN101 · MS-COCO partial labels 10%mAP: 70.4 -> 65.9Hurt
fine-tune the aggregation function (ablation)DualCoOp++ · MS-COCO partial labels 10%mAP: 81.4 -> 81.9Improved

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