AI Paper Atlas

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

Visual Clues: Bridging Vision and Language Foundations for Image Paragraph Captioning

Yujia Xie, Luowei Zhou, Xiyang Dai, Lu Yuan, Nguyen Bach, Ce Liu, Michael Zeng

NeurIPS 2022 · Vision-language and multimodal › CLIP and zero-shot · approach: Open-vocabulary alignment from VLMs · 9 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
Use structured visual clues with VG-domain tags instead of BLIP-largeImage paragraph captioning · Stanford dataset testSPIPE F-score: 7.6 -> 10.0Improved
Use structured visual clues with VG-domain tags instead of Socratic modelImage paragraph captioning · Stanford dataset testSPIPE F-score: 3.2 -> 10.0Improved
Specify the tag vocabulary to the VG domain rather than using general-domain tagsBEST · Stanford dataset testSPIPE F-score: 8.8 -> 10.0Improved
Use visual clues rather than Socratic prompts for VQAVisual question answering · GQA (Generative)accuracy: 24.95 -> 37.00Improved
Use visual clues rather than Socratic prompts for VQAVisual question answering · GQA (Discriminative)accuracy: 26.89 -> 39.93Improved
Use visual clues rather than Socratic prompts for VQAVisual question answering · OK-VQA (Generative)accuracy: 16.50 -> 28.89Improved
Add local object descriptions (ablation)BEST-VG domain · Stanford dataset testSPIPE F-score: 8.0 -> 10.0Improved
Add the caption model (ablation)BEST-VG domain · Stanford dataset testSPIPE F-score: 8.7 -> 10.0Improved
Add local descriptions and captions alongside tags (ablation)BEST-VG domain · Stanford dataset testSPIPE F-score: 5.9 -> 10.0Improved
Use the open-vocabulary extractor rather than YOLOv5 (ablation)BEST-VG domain · Stanford dataset testSPIPE F-score: 9.0 -> 10.0Improved
Use Florence-H rather than CLIP ViT-L/14 as the tagger (ablation)BEST-VG domain · Stanford dataset testSPIPE F-score: 7.8 -> 10.0Improved
Use Davinci-text-001 rather than the smaller Curie language model (ablation)BEST-VG domain · Stanford dataset testSPIPE F-score: 8.9 -> 10.0Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 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