AI Paper Atlas

Research map / Video / Video-language understanding

Exploring Group Video Captioning with Efficient Relational Approximation

Wang Lin, Tao Jin, Ye Wang, Wenwen Pan, Linjun Li, Xize Cheng, Zhou Zhao

ICCV 2023 · Video › Video-language understanding · approach: Encoder-decoder video captioning · 7 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 ERA instead of traversal for shared content identificationgroup video captioning model · YouCook2 Captionstraining time per epoch: 3884.4s -> 17.6sImproved
use ERA instead of traversal on caption qualitygroup video captioning model · YouCook2 CaptionsCIDEr: 179.4 -> 180.8No clear effect
use ERA instead of simple averaging of video featuresgroup video captioning model · YouCook2 CaptionsCIDEr: +10.2% CIDEr (170.6 -> 180.8)Improved
add contextual feature refinery instead of subtraction (ablation)ERA captioner · YouCook2 CaptionsCIDEr: 168.3 -> 178.7Improved
add contextual features (subtraction) to ERA captioner (ablation)ERA captioner · YouCook2 CaptionsCIDEr: 161.7 -> 168.3Improved
add intra-group contrastive loss (IGC) on top of refinery (ablation)ERA captioner · YouCook2 CaptionsCIDEr: 178.7 -> 180.8Improved
add IGC on top of refinery for ERA (ablation)ERA captioner · YouCook2 CaptionsWER: 75.42 -> 75.77No clear effect
add IGC on top of refinery for Traversal (ablation)Traversal captioner · ActivityNet CaptionsBLEU4: 17.65 -> 16.53Mixed
remove reference videos from the group (ablation)ERA captioner · YouCook2 CaptionsCIDEr: 180.8 -> 69.81Hurt
reduce number of target videos from 3 to 1 (ablation)ERA captioner · YouCook2 CaptionsCIDEr: 180.8 -> 163.1Hurt

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