AI Paper Atlas

Research map / Detection and segmentation / Tracking

Bridging Images and Videos: A Simple Learning Framework for Large Vocabulary Video Object Detection

Sanghyun Woo, Kwanyong Park, Seoung Wug Oh, In So Kweon, Joon-Young Lee

ECCV 2022 · Detection and segmentation › Tracking · approach: Memory and correlation-based MOT · 7 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 SimLearn unified learningFasterRCNN-RFS* · TAO-valTrack AP50: 13.4 -> 19.6Improved
add SimLearn unified learningCenterNet2-RFS · TAO-valTrack AP50: 18.9 -> 21.6Improved
add SimLearn unified learningFasterRCNN-Seesaw · TAO-valTrack AP50: 15.4 -> 20.2Improved
use strong zoom-in/out scale jittering (0.1-2.0) instead of small range (ablation)LVIS-tracker · TAO-valTrack AP50: 13.2 -> 14.4Improved
add mosaicing to zoom in/out (ablation)LVIS-tracker · TAO-valTrack AP50: 14.4 -> 17.8Improved
add mosaicing augmentation alone (ablation)LVIS-tracker · TAO-valBox AP: 19.0 -> 16.2Hurt
add pseudo-label distillation and negative correction to vanilla teacher-student (ablation)Teacher-student · TAO-valTrack AP50: 15.9 -> 18.5Improved
use soft pseudo labels instead of hard (ablation)Teacher-student · TAO-valTrack AP50: 15.9 -> 18.5Improved
use MSE instead of KL distillation loss (ablation)Teacher-student · TAO-valTrack AP50: 16.8 -> 18.5Improved
add semantic consistency loss (ablation)Teacher-student tracker · TAO-valTrack AP50: 18.5 -> 19.5Improved
avoid naive fine-tuning on TAO by using teacher-student (ablation)LVIS-tracker · TAO-valTrack AP50: 11.4 -> 18.5Improved
use teacher-student instead of track-only fine-tuningLVIS to TAO transfer · TAO-val NEW classesTrack AP50: 16.9 -> 20.6Improved

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