AI Paper Atlas

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

Quilt-1M: One Million Image-Text Pairs for Histopathology

Wisdom Oluchi Ikezogwo, Mehmet Saygin Seyfioglu, Fatemeh Ghezloo, Dylan Stefan Chan Geva, Fatwir Sheikh Mohammed, Pavan Kumar Anand, Ranjay Krishna, Linda Shapiro

NeurIPS 2023 · Vision-language and multimodal › CLIP and zero-shot · approach: Recaptioning and text-supervised pretraining · 55 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
fine-tune CLIP ViT-B/32 on QUILT-1M (QuiltNet) instead of using PLIP (trained on OpenPath)CLIP ViT-B/32 · SICAPv2 linear probe 100%accuracy %: 73.0 -> 75.48Improved
fine-tune CLIP ViT-B/32 on QUILT-1M instead of using PLIPCLIP ViT-B/32 · SkinCancer linear probe 100%accuracy %: 93.29 -> 93.03No clear effect
fine-tune CLIP on QUILT-1M (ViT-B/32) vs off-the-shelf CLIPCLIP ViT-B/32 · NCT-CRC linear probe 100%accuracy %: 91.83 -> 95.22Improved
QUILT-1M-tuned model vs BiomedCLIP (PubMedBERT text encoder)ViT-B/16 PMB/256 · QUILT-1M holdout text-to-image R@50R@50 %: 14.99 -> 30.28Improved
QUILT-1M-tuned model vs BiomedCLIP on ARCHViT-B/16 PMB/256 · ARCH text-to-image R@50R@50 %: 53.24 -> 55.14Improved
train CLIP from scratch instead of fine-tuning pre-trained CLIP on QUILT-1M (ablation)QuiltNet · downstream histopathology tasksperformanceHurt
use 10% balanced subset instead of 100% training data for linear probelinear probe · imbalanced datasetsaccuracy %Mixed

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