AI Paper Atlas

Research map / Video / Video-language understanding

VLG: General Video Recognition with Web Textual Knowledge

Jintao Lin, Zhaoyang Liu, Wenhai Wang, Wayne Wu, Limin Wang

IJCV 2024 · Video › Video-language understanding · approach: Video-language pretraining · 5 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 video-language pretraining (VLP) (ablation)VLG ViT-B/16 · Kinetics-Close ablationtop-1 accuracy: 76.0 -> 81.8Improved
initialize with CLIP pretrained weights instead of random (ablation)VLG ViT-B/16 · Kinetics-Close ablationtop-1 accuracy: 32.6 -> 81.8Improved
use bi-modal attention head instead of FC head (ablation)VLG ViT-B/16 · Kinetics-Close ablationtop-1 accuracy: 79.5 -> 81.8Improved
use bi-modal attention head instead of KNN head (ablation)VLG ViT-B/16 · Kinetics-Close ablationtop-1 accuracy: 79.9 -> 81.8Improved
use text selection ruler (TSR) instead of random sentence selection (ablation)VLG ViT-B/16 · Kinetics-Close ablationtop-1 accuracy: 80.0 -> 81.8Improved
use TSR salient sentences instead of basic prompts only (ablation)VLG ViT-B/16 · Kinetics-Close ablationtop-1 accuracy: 78.9 -> 81.8Improved
increase temporal module layers from 0 to 6 (ablation)VLG ViT-B/16 · Kinetics-Closetop-1 accuracy: 79.8 -> 81.8Improved
increase temporal module layers from 6 to 8 (ablation)VLG ViT-B/16 · Kinetics-Closetop-1 accuracy: 81.8 -> 80.6Hurt
add temporal module layers (0 to 8) in few-shot setting (ablation)VLG ViT-B/16 · Kinetics-Fewshottop-1 accuracy: 96.9 -> 95.2Hurt
add logits distillation with 1-alpha=0.5 (ablation)VLG ResNet-50 · Kinetics-LToverall top-1 accuracy: 57.5 -> 60.8Improved
raise logits distillation weight 1-alpha to 0.9 (ablation)VLG ResNet-50 · Kinetics-LToverall top-1 accuracy: 60.8 -> 54.7Hurt
use feature distillation (1-alpha=0.5) instead of logits distillation (ablation)VLG ResNet-50 · Kinetics-LToverall top-1 accuracy: 60.8 -> 59.7Hurt

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