AI Paper Atlas

Research map / Video / Video segmentation and tracking

Video Mask Transfiner for High-Quality Video Instance Segmentation

Lei Ke, Henghui Ding, Martin Danelljan, Yu-Wing Tai, Chi-Keung Tang, Fisher Yu

ECCV 2022 · Video › Video segmentation and tracking · approach: Mask propagation for unsupervised video segmentation · 24 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
use quadtree sequence grouping across frames (clip length 5) (ablation)VMT · HQ-YTVIS valAP^B: 30.2 -> 31.4Improved
use quadtree sequence grouping across frames (all frames) (ablation)VMT · HQ-YTVIS valAP^B: 32.3 -> 33.7Improved
lengthen the clip from 1 frame to all frames (ablation)VMT · HQ-YTVIS valAP^B: 26.1 -> 33.7Improved
use video-level instead of frame-level instance queries in the detector (ablation)VMT · HQ-YTVIS valAP^B: 31.8 -> 33.7Improved
use video-level instance query instead of fixed FCN detector (ablation)VMT · HQ-YTVIS valAP^B: 32.2 -> 33.7Improved
refine 3D incoherent regions instead of detected object boundaries (ablation)VMT · HQ-YTVIS valAP^B: 32.8 -> 33.7Improved
train on iteratively self-corrected HQ-YTVIS instead of YTVISVMT · HQ-YTVIS valAP^B: 30.5 -> 33.7Improved
train on HQ-YTVIS instead of YTVISSeqFormer · HQ-YTVIS valAP^B: 28.9 -> 31.0Improved
train on HQ-YTVIS instead of YTVISMaskTrack · HQ-YTVIS valAP^B: 19.8 -> 21.7Improved
train on HQ-YTVIS instead of YTVISVMT · OVIS valAP^M: 15.9 -> 17.1Improved
iterative self-correction with VMT instead of MaskTrack or SeqFormer (ablation)VMT · HQ-YTVIS valAP^B: around 60.0 for baselinesImproved
replace SeqFormer with VMT on OVIS (Swin-L)SeqFormer · OVIS valAP: 17.9 -> 19.8Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 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