AI Paper Atlas

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

UniCLIP: Unified Framework for Contrastive Language-Image Pre-training

Janghyeon Lee, Jongsuk Kim, Hyounguk Shon, Bumsoo Kim, Seung Hwan Kim, Honglak Lee, Junmo Kim

NeurIPS 2022 · Vision-language and multimodal › CLIP and zero-shot · approach: CLIP training recipes and data curation · 21 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 augmentation embedding to the image projection head (ablation)UniCLIP (3 ResBlocks) · ImageNet-1k zero-shot, CC3M 50 ep, ViT-B/16accuracy: 24.46 -> 27.84Improved
use residual blocks instead of MLP head with augmentation embedding (ablation)UniCLIP · ImageNet-1k zero-shot, CC3M 50 ep, ViT-B/16accuracy: 24.15 -> 27.84Improved
use 3 ResBlocks instead of 1 ResBlock (ablation)UniCLIP · ImageNet-1k zero-shot, CC3M 50 ep, ViT-B/16accuracy: 27.67 -> 27.84No clear effect
replace 3 weak augmentations by 1 weak + 2 strong (with augmentation embedding) (ablation)UniCLIP · ImageNet-1k zero-shot, CC3M 50 ep, ViT-B/16accuracy: 23.40 -> 27.84Improved
use 3 strong augmentations without augmentation embedding instead of 1 weak 2 strong (ablation)UniCLIP · ImageNet-1k zero-shot, CC3M 50 ep, ViT-B/16accuracy: 24.76 -> 22.60Hurt
domain-dependent temperature and offset instead of shared (ablation)UniCLIP · ImageNet-1k zero-shot, CC3M 50 ep, ViT-B/16accuracy: 25.51 -> 27.84Improved
unified supervision instead of separated supervision (ablation)UniCLIP · ImageNet-1k zero-shot, CC3M 50 ep, ViT-B/16accuracy: 26.59 -> 27.84Improved
MP-NCE instead of SupCon loss (ablation)UniCLIP · ImageNet-1k zero-shot, CC3M 50 ep, ViT-B/16accuracy: 23.04 -> 27.84Improved
SupCon instead of MIL-NCE (ablation)UniCLIP · ImageNet-1k zero-shot, CC3M 50 ep, ViT-B/16accuracy: 22.23 -> 23.04Improved
add domain balancing weights w to MP-NCE (ablation)UniCLIP · ImageNet-1k zero-shot, CC3M 50 ep, ViT-B/16accuracy: 26.41 -> 27.84Improved
add trivial pair (z_i, z_i) to positive set (ablation)UniCLIP MP-NCE · ImageNet-1k zero-shot, CC3M 50 ep, ViT-B/16accuracy: 24.60 -> 26.41Improved
UniCLIP versus SLIP and DeCLIP on pre-trainingViT-B/32 YFCC15M · 11-dataset linear probingaverage accuracy: 71.8 -> 75.2Improved

Later papers that compared with this method

1 later accepted paper reports UniCLIP as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

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