Research map / Label-efficient and robust learning / Noisy labels
CAPro: Webly Supervised Learning with Cross-modality Aligned Prototypes
Yulei Qin, Xingyu Chen, Yunhang Shen, Chaoyou Fu, Yun Gu, Ke Li, Xing Sun, Rongrong Ji
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add CAPro cross-modality aligned prototypes and collective bootstrapping over vanilla | R50 vanilla · WebVision1k | top-1 accuracy: 72.6 -> 74.2 | Improved |
| add CAPro over vanilla | R50 vanilla · ImageNet1k | top-1 accuracy: 67.0 -> 68.0 | Improved |
| add CAPro over vanilla | R50 vanilla · NUS-WIDE (Web) | mAP: 38.3 -> 48.0 | Improved |
| add CAPro over vanilla | R50 vanilla · Open-set recognition, WebVision | C-F1: 54.6 -> 62.2 | Improved |
| add text encoding plus the paper's text enhancement (MiniLM) (ablation) | CAPro · Google500 | top-1 accuracy: 71.5 -> 75.5 | Improved |
| replace VSGraph text enhancement with the paper's enhancement (MiniLM) (ablation) | CAPro · ImageNet500 | top-1 accuracy: 66.9 -> 71.5 | Improved |
| replace MiniLM text encoder with XLNet (ablation) | CAPro with text enhancement · Google500 | top-1 accuracy: 75.5 -> 75.4 | No clear effect |
| replace MiniLM text encoder with GPT-Neo (1.3B) (ablation) | CAPro with VSGraph enhancement · Google500 | top-1 accuracy: 72.0 -> 72.0 | No clear effect |
| add collective bootstrapping as reference provider (ablation) | CAPro with text enhancement · Google500 | top-1 accuracy: 75.5 -> 76.0 | Improved |
| use SCC as reference provider instead of CB (ablation) | CAPro · Google500 | top-1 accuracy: 75.5 -> 73.8 | Hurt |
| use NCR as reference provider instead of CB (ablation) | CAPro · NUS-WIDE | O-F1: 45.4 -> 43.4 | Hurt |
| add mix-up on top of collective bootstrapping (ablation) | CAPro · Google500 | top-1 / top-5 accuracy: 76.0 -> 76.5 top1; 91.3 -> 91.1 top5 | Mixed |
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.
- MoPro — MoPro: Webly Supervised Learning with Momentum Prototypes (ICLR 2021) · better on 1 of 1 shared result
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
- T-MARS: Improving Visual Representations by Circumventing Text Feature Learning (ICLR 2024 · similarity 0.85)
- Mind the Gap Between Prototypes and Images in Cross-domain Finetuning (NeurIPS 2024 · similarity 0.85)
- Toward Modality Gap: Vision Prototype Learning for Weakly-supervised Semantic Segmentation with CLIP (AAAI 2025 · similarity 0.85)
- CovMatch: Cross-Covariance Guided Multimodal Dataset Distillation with Trainable Text Encoder (NeurIPS 2025 · similarity 0.85)
- On the Value of Cross-Modal Misalignment in Multimodal Representation Learning (NeurIPS 2025 · similarity 0.85)
- Visual-Text Cross Alignment: Refining the Similarity Score in Vision-Language Models (ICML 2024 · similarity 0.84)
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.
