Research map / Vision-language and multimodal / Human and embodied modeling
RedCaps: Web-curated image-text data created by the people, for the people
Karan Desai, Gaurav Kaul, Zubin Trivadi Aysola, Justin Johnson
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 |
|---|---|---|---|
| Train VirTex-v2 on Reddit image-caption data instead of CC-3M | VirTex-v2 visual representations · Zero-shot classification across six of seven datasets | top-1 accuracy: RedCaps outperforms CC-3M on six out of seven datasets | Improved |
| Train on the 2020 RedCaps subset rather than the full RedCaps dataset (ablation) | VirTex-v2 · Zero-shot classification | top-1 accuracy: RedCaps-20 vs RedCaps: 41.8 vs 42.4 (Pets); 54.6 vs 53.8 (Food); 33.5 vs 26.2 (Flowers); 3.2 vs 3.1 (Cars); 2.3 vs 3.6 ( | Mixed |
| Train VirTex-v2 on Reddit image-caption data instead of CC-3M | VirTex-v2 visual representations · ImageNet | k-NN accuracy (k=20): 45.4 vs 52.0 | Improved |
| Train VirTex-v2 on Reddit image-caption data instead of CC-3M | VirTex-v2 visual representations · ImageNet | zero-shot top-1 accuracy: 20.7 vs 22.7 | Improved |
| Train VirTex-v2 on Reddit image-caption data instead of CC-3M | VirTex-v2 visual representations · ImageNet | linear classification accuracy: 53.9 vs 53.4 | Hurt |
| Train VirTex-v2 on Reddit image-caption data instead of CC-3M | VirTex-v2 visual representations · COCO instance segmentation | segmentation AP: 37.2 vs 37.0 | No clear effect |
| Train the captioning model on RedCaps instead of CC-3M | VirTex-v2 captioner · Held-out test set of 1K images | images where workers preferred the generated caption: 633/1000 | Improved |
Closest papers by meaning
- A Picture is Worth More Than 77 Text Tokens: Evaluating CLIP-Style Models on Dense Captions (CVPR 2024 · similarity 0.88)
- CapsFusion: Rethinking Image-Text Data at Scale (CVPR 2024 · similarity 0.87)
- Benchmarking Large Vision-Language Models via Directed Scene Graph for Comprehensive Image Captioning (CVPR 2025 · similarity 0.85)
- What If We Recaption Billions of Web Images with LLaMA-3? (ICML 2025 · similarity 0.85)
- OmniCorpus: A Unified Multimodal Corpus of 10 Billion-Level Images Interleaved with Text (ICLR 2025 · similarity 0.85)
- Panoptic Captioning: An Equivalence Bridge for Image and Text (NeurIPS 2025 · similarity 0.85)
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.
