AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

BIT: Matching-based Bi-directional Interaction Transformation Network for Visible-Infrared Person Re-Identification

Haoxuan Xu, Guanglin Niu

CVPR 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Cross-modality and occlusion re-identification · 1 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 Bi-directional Cross Interaction (BCI) module to the PMT baseline (ablation)PMT baseline · SYSU-MM01 All-Search Single-ShotRank-1 / mAP (%): 69.23 -> 75.24 Rank-1; 66.02 -> 73.35 mAPImproved
add Aggregation Contrastive Loss LAC after BCI (ablation)BIT · SYSU-MM01 All-Search Single-ShotRank-1 / mAP (%): 75.24 -> 76.42 Rank-1; 73.35 -> 74.54 mAPImproved
add Query Aware Scoring module (ablation)BIT · SYSU-MM01 All-Search Single-ShotRank-1 / mAP (%): 75.24 -> 79.53 Rank-1; 73.35 -> 79.02 mAPImproved
add LAC on top of BCI and QA Scoring (full model) (ablation)BIT · SYSU-MM01 All-Search Single-ShotRank-1 (%): 79.53 -> 80.53Improved
use standard cross-attention in place of BCI (ablation)PMT baseline · SYSU-MM01 All-Search Single-ShotRank-1 / mAP (%): 69.23 -> 68.68 Rank-1; 66.02 -> 64.67 mAPHurt
use bi-directional BCI instead of standard cross-attention (ablation)PMT baseline · SYSU-MM01 All-Search Single-ShotRank-1 (%): 68.68 -> 75.24Improved
set Top-k to 3 (ablation)QA Scoring · SYSU-MM01 All-Search Single-ShotRank-1 (%)Improved
set penalty coefficient alpha to 0.2 and loss weight lambda to 0.6 (ablation)BIT · SYSU-MM01 All-Search Single-ShotRank-1 (%)Improved
add re-ranking to BITBIT · SYSU-MM01 All-Search Single-ShotRank-1 (%): 80.53 -> 84.42Improved
use coarse candidate filtering at inference with K=50BIT · SYSU-MM01 inference latencylatency per query: 10ms (PMT) vs 14ms (BIT)No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

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