AI Paper Atlas

Research map / Video / Video segmentation and tracking

Video Instance Segmentation via Multi-Scale Spatio-Temporal Split Attention Transformer

Omkar Thawakar, Sanath Narayan, Jiale Cao, Hisham Cholakkal, Rao Muhammad Anwer, Muhammad Haris Khan, Salman Khan, Michael Felsberg, Fahad Shahbaz Khan

ECCV 2022 · Video › Video segmentation and tracking · approach: Mask propagation for unsupervised video segmentation · 22 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 MS-STS attention module to the encoder (ablation)SeqFormer baseline · YouTube-VIS 2019 val (ablation)AP: 46.4 -> 48.4Improved
add temporal attention block in the decoder (ablation)SeqFormer + MS-STS · YouTube-VIS 2019 val (ablation)AP: 48.4 -> 49.1Improved
add adversarial fg-bg separability loss (ablation)SeqFormer + MS-STS + T-Dec · YouTube-VIS 2019 val (ablation)AP: 49.1 -> 50.1Improved
add all three contributions to SeqFormer (ablation)SeqFormer · YouTube-VIS 2019 valAP: 46.4 -> 50.1Improved
use MS-STS VIS instead of SeqFormer on fast motion videosSeqFormer · YouTube-VIS 2021 new videos, fast motionAP: 48.9 -> 54.2Improved
use MS-STS VIS instead of SeqFormer on size-change videosSeqFormer · YouTube-VIS 2021 new videos, size changeAP: 49.5 -> 53.7Improved
use MS-STS VIS instead of SeqFormer on aspect-ratio change videosSeqFormer · YouTube-VIS 2021 new videos, aspect-ratio changeAP: 49.2 -> 53.9Improved
use MS-STS VIS instead of SeqFormer with Swin-LSeqFormer · YouTube-VIS 2019 val (Swin-L)AP: 59.3 -> 61.0Improved
use MS-STS VIS instead of SeqFormer at AP75SeqFormer · YouTube-VIS 2019 val (ResNet-50)AP75: 51.8 -> 56.6Improved
use MS-STS VIS instead of SeqFormerSeqFormer · YouTube-VIS 2021 valAP: 40.5 -> 42.2Improved

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