AI Paper Atlas

Research map / Video / Video segmentation and tracking

Cross-Platform Video Person ReID: A New Benchmark Dataset and Adaptation Approach

Shizhou Zhang, Wenlong Luo, De Cheng, Qingchun Yang, Lingyan Ran, Yinghui Xing, Yanning Zhang

ECCV 2024 · Video › Video segmentation and tracking · approach: Spatial-temporal video person re-identification · 28 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 visual-semantic alignment (CLIP text tokens) to full fine-tuning (ablation)FT-CLIP · LS-VIDmAP: 76.10 -> 84.07Improved
add visual-semantic alignment to the Intra-Frame Adapter (ablation)IFA · LS-VIDmAP: 77.31 -> 84.16Improved
add Cross-Frame Attention Adapter (set-level modeling) (ablation)IFA+VSA · LS-VIDmAP: 84.16 -> 85.20Improved
add Platform-Bridge Prompt (ablation)VSLA-CLIP · G2A-VReIDmAP: 79.70 -> 81.29Improved
replace full fine-tuning with Intra-Frame Adapter (4.7M vs 86.1M tunable) (ablation)CLIP image encoder · LS-VIDmAP: 76.10 -> 77.31Improved
use VSLA-CLIP adapter instead of full fine-tuning FT-CLIPViFi-CLIP initialized model · G2A-VReIDmAP: 78.11 -> 79.70Improved
increase adapter width alpha from 64 to 256 (ablation)VSLA · LS-VIDmAP: 79.58 -> 85.20Improved
increase adapter width alpha from 256 to 384 (ablation)VSLA · LS-VIDmAP: 85.20 -> 85.09No clear effect
use all three alignment losses (v2sce, i2t, t2i) rather than one (ablation)VSLA-CLIP · LS-VIDmAP: 84.73 -> 85.20Improved
initialize from ViFi-CLIP weights instead of OpenAI CLIP (ablation)VSLA-CLIP · LS-VIDmAP: 84.05 -> 85.20Improved
use temporal modeling video ReID methods versus stripe-alignment methods on aerial databaselines · G2A-VReIDmAP: AP3D 67.7 vs MGH 76.7Hurt

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