AI Paper Atlas

Research map / Detection and segmentation / Tracking

Heterogeneous Diversity Driven Active Learning for Multi-Object Tracking

Rui Li, Baopeng Zhang, Jun Liu, Wei Liu, Jian Zhao, Zhu Teng

ICCV 2023 · Detection and segmentation › Tracking · approach: Association cues for online MOT · 13 citations (OpenAlex)

Read the paper

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 multi-frame cooperation to IFSN (ablation)HD-AMOT · MOT17 (20% budget)MOTA / IDF1: MOTA 62.76 -> 63.23, IDF1 65.52 -> 66.47Improved
use cluster-based batch sampling instead of top-N score batch sampling (ablation)HD-AMOT · MOT17 (20% budget)MOTA / IDF1: MOTA 62.18 -> 63.23, IDF1 64.55 -> 66.47Improved
batch sampling instead of one frame per iteration (ablation)HD-AMOT · MOT17 (20% budget)MOTA / IDF1: MOTA 63.07 -> 63.23, IDF1 65.62 -> 66.47Improved
add novelty histograms to action representation (ablation)HD-AMOT · MOT17 (20% budget)MOTA / IDF1: MOTA 61.25 -> 63.23, IDF1 63.41 -> 66.47Improved
add geometric topology to state (ablation)HD-AMOT · MOT17 (20% budget)MOTA: 61.73 -> 63.23Improved
add geometric scale/topology histogram HGh to action (ablation)HD-AMOT · MOT17 (20% budget)IDF1: 66.61 (without) vs 66.47 (full)No clear effect
use HD-AMOT sampling instead of uniform samplingFairMOT · MOT17 testIDF1: MOTA 70.40 -> 71.24, IDF1 70.05 -> 69.87Mixed
use HD-AMOT instead of random samplingFairMOT · MOT20 testMOTA: 58.21 -> 59.24Improved
apply Coreset/CDAL image-level active learning directly to MOTFairMOT · MOT20 testMOTA: Random 58.21 -> Coreset 56.66Hurt
train with 50% actively selected labels instead of full supervisionFairMOT · MOT15MOTA: gain of 2% MOTA and 0.8% IDF1Improved

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