AI Paper Atlas

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

Learning Generalisable Omni-Scale Representations for Person Re-Identification

Kaiyang Zhou, Yongxin Yang, Andrea Cavallaro, Tao Xiang

TPAMI 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Supervised person re-identification · 381 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 omni-scale streams T=4 instead of single-scale T=1 (ablation)OSNet · MSMT17Rank-1: 52.9 -> 71.0Improved
use unified aggregation gate instead of addition (ablation)OSNet · MSMT17Rank-1: 64.5 -> 71.0Improved
use unified aggregation gate instead of concatenation (ablation)OSNet · MSMT17Rank-1: 66.4 -> 71.0Improved
share one aggregation gate across streams instead of separate AGs (ablation)OSNet · MSMT17Rank-1: 68.3 -> 71.0Improved
use channel-wise gates instead of stream-wise gates (ablation)OSNet · MSMT17Rank-1: 67.9 -> 71.0Improved
use dynamic input-dependent gates instead of learned-and-fixed gates (ablation)OSNet · MSMT17Rank-1: 64.5 -> 71.0Improved
use heterogeneous-depth streams instead of same-depth ResNeXt-like streams (ablation)OSNet · CUHK03Rank-1: 54.4 -> 57.1Improved
use Lite 3x3 factorised convolutions instead of standard convolutions (ablation)OSNet · MSMT17Rank-1: 70.5 -> 71.0No clear effect
add instance normalisation via architecture search (OSNet-AIN vs OSNet)OSNet · Market1501->DukeRank-1: 44.7 -> 52.4Improved
add instance normalisation via architecture search (OSNet-AIN vs OSNet)OSNet · Duke->Market1501Rank-1: 52.2 -> 61.0Improved
search IN placement instead of hand-placing IN in lowest layers (ablation)OSNet-IN · Market1501->DukeRank-1: 47.9 -> 52.4Improved
search IN placement instead of hand-placing IN in lowest layers (ablation)OSNet-IN · Duke->Market1501Rank-1: 57.8 -> 61.0Improved

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