AI Paper Atlas

Research map / Video / Video segmentation and tracking

SeqFormer: Sequential Transformer for Video Instance Segmentation

Junfeng Wu, Yi Jiang, Song Bai, Wenqing Zhang, Xiang Bai

ECCV 2022 · Video › Video segmentation and tracking · approach: Matching-based video object segmentation · 168 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 instance query decomposition into frame-level box queries (ablation)SeqFormer · YouTube-VIS 2019 valAP: 34.1 -> 45.1Improved
keep spatial and temporal dimensions separate instead of flattening them (ablation)SeqFormer · YouTube-VIS 2019 valAP: 35.1 -> 42.5Improved
add multi-scale feature maps (ablation)SeqFormer (single-scale) · YouTube-VIS 2019 valAP: 42.5 -> 45.1Improved
use weighted-sum temporal aggregation instead of sum (ablation)SeqFormer · YouTube-VIS 2019 valAP: 30.6 -> 45.1Improved
use weighted-sum temporal aggregation instead of average (ablation)SeqFormer · YouTube-VIS 2019 valAP: 43.2 -> 45.1Improved
generate mask head from all frames instead of 1 frame (ablation)SeqFormer · YouTube-VIS 2019 valAP: 38.1 -> 45.1Improved
use 5 frames instead of all frames for the instance representation (ablation)SeqFormer · YouTube-VIS 2019 valAP: 44.6 -> 45.1No clear effect
train with COCO pseudo videos (joint training) (ablation)SeqFormer · YouTube-VIS 2019 valAP: 45.1 -> 47.4Improved
evaluate clip-wise with clip matching instead of whole videoSeqFormer · YouTube-VIS 2019 valAP: 45.1 -> 42.0Hurt
use a stronger Swin-L backbone instead of ResNet-50SeqFormer · YouTube-VIS 2019 valAP: 47.4 -> 59.3Improved
use SeqFormer instead of IFCSeqFormer · YouTube-VIS 2021 valAP: 36.6 -> 40.5Improved

Later papers that compared with this method

6 later accepted papers report SeqFormer as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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