AI Paper Atlas

Research map / Video / Video segmentation and tracking

CTVIS: Consistent Training for Online Video Instance Segmentation

Kaining Ying, Qing Zhong, Weian Mao, Zhenhua Wang, Hao Chen, Lin Yuanbo Wu, Yifan Liu, Chengxiang Fan, Yunzhi Zhuge, Chunhua Shen

ICCV 2023 · Video › Video segmentation and tracking · approach: Mask propagation for unsupervised video segmentation · 80 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 memory bank, momentum-averaged embeddings and noise (full CTVIS) to single-reference baseline (ablation)CTVIS (ResNet-50) · YTVIS19AP: 51.6 -> 55.1Improved
use momentum-averaged embeddings in the memory bank to build contrastive items (ablation)CTVIS · YTVIS19AP: 52.1 -> 54.2Improved
add noise to memory bank updates (ablation)CTVIS · YTVIS19AP: 54.2 -> 55.1Improved
add noise to memory bank updates (ablation)CTVIS · OVISAP: 34.9 -> 35.5Improved
train with more frames per video (up to 8 or 10) using CTVIS contrastive items (ablation)CTVIS · OVISAP: 31.2 -> 35.5Improved
train IDOL with multiple reference frames and more training frames (ablation)IDOL multi-reference (Mask2Former) · OVISAP: 31.2 -> 27.3Hurt
add CTVIS to IDOL with Deformable DETR (ablation)IDOL (Deformable DETR) · YTVIS19AP: 49.5 -> 53.7Improved
use supplementary background negatives only (no major negatives) (ablation)CTVIS · YTVIS19AP: 55.1 -> 16.5Hurt
sample supplementary negatives from preceding frame (local) instead of all previous frames (global) (ablation)CTVIS · OVISAP: 33.4 -> 35.5Improved
add copy&paste and crop augmentation to rotation for pseudo-videos (ablation)CTVIS pseudo-video training · OVIS*AP: 27.3 -> 30.5Improved
add crop to rotation for pseudo-videos (ablation)CTVIS pseudo-video training · YTVIS21*AP: 48.5 -> 48.7No clear effect
train CTVIS on pseudo-videos from 5% of frames instead of MinVIS on 100%CTVIS · OVISAP: 39.4 -> 41.9Improved

Later papers that compared with this method

7 later accepted papers report CTVIS as a baseline in their results tables; 7 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 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