AI Paper Atlas

Research map / Video / Video-language understanding

R^2-Tuning: Efficient Image-to-Video Transfer Learning for Video Temporal Grounding

Ye Liu, Jixuan He, Wanhua Li, Junsik Kim, Donglai Wei, Hanspeter Pfister, Chang Wen Chen

ECCV 2024 · Video › Video-language understanding · approach: Temporal video grounding

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 video-level contrastive constraint (ablation)R2-Tuning · QVHighlights valMR mAP: 44.01 -> 46.74Improved
add layer-wise contrastive constraint (ablation)R2-Tuning · QVHighlights valMR mAP: 44.01 -> 44.60Improved
combine video-level and layer-wise constraints (ablation)R2-Tuning · QVHighlights valMR mAP: 44.01 -> 47.59Improved
refine features in reversed (back-to-front) order instead of forward (ablation)R2-Tuning · QVHighlights valMR mAPImproved
fuse multi-layer CLIP features (mean of last K layers) instead of a single layer (ablation)R2 Block · QVHighlights valMR mAPImproved
cross-attention before self-attention instead of self before cross (ablation)R2 Block · QVHighlights valMR mAP: 45.21 -> 47.59Improved
share R2 Block parameters across refinement steps (ablation)R2-Tuning · QVHighlights valMR mAP: 47.48 -> 47.59No clear effect
increase temporal transformer layers from 1 to 6 (ablation)R2 Block · QVHighlights valMR mAP: 47.59 -> 47.42No clear effect
use R2-Tuning instead of full fine-tuning of CLIPCLIP ViT-B/32 · QVHighlights valMR mAP: 42.41 -> 47.59Improved
full fine-tuning of CLIP versus frozen CLIPCLIP ViT-B/32 · QVHighlights valMR mAP: 42.22 -> 42.41No clear effect
use LoRA on CLIP instead of frozen CLIPCLIP ViT-B/32 · QVHighlights valMR mAP: 42.22 -> 41.29Hurt
use R2-Tuning without video backbone against UniVTG with SlowFast and post pre-trainingQVHighlights test · QVHighlights testMR mAP avg: 43.63 -> 46.17Improved

Later papers that compared with this method

6 later accepted papers report R2-Tuning as a baseline in their results tables; 6 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