AI Paper Atlas

Research map / Video / Video segmentation and tracking

URVOS: Unified Referring Video Object Segmentation Network with a Large-Scale Benchmark

Seonguk Seo, Joon-Young Lee, Bohyung Han

ECCV 2020 · Video › Video segmentation and tracking · approach: Matching-based video object segmentation · 202 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 cross-modal attention (ablation)URVOS · Refer-Youtube-VOS valJ: 38.88 -> 45.27Improved
add memory attention (ablation)URVOS · Refer-Youtube-VOS valJ: 39.38 -> 45.27Improved
add cross-modal attention (ablation)URVOS · Refer-DAVIS17 valJ: 38.25 -> 47.29Improved
add memory attention (ablation)URVOS · Refer-DAVIS17 valJ: 39.43 -> 47.29Improved
add GRU over multi-frame features to image-based baselineBaseline · Refer-Youtube-VOS valJ: 33.34 -> 34.79Improved
use two-stage inference with anchor frame and confidence-selected memory instead of forward propagation (ablation)URVOS · Refer-Youtube-VOS valJ: 43.13 -> 45.27Improved
select memory frames by confidence instead of previous T frames (ablation)URVOS · Refer-Youtube-VOS valJ: 44.58 -> 45.27Improved
add second-stage mask propagation (ablation)URVOS · Refer-Youtube-VOS valJ: 41.34 -> 45.27Improved
iterate second stage up to five times (ablation)URVOS · Refer-Youtube-VOS valJ: 45.27 -> 45.44No clear effect
increase training data from 10% to 100% of Refer-Youtube-VOS (ablation)URVOS · Refer-Youtube-VOS valJ: 30.73 -> 45.27Improved
pretrain on Refer-Youtube-VOS instead of RefCOCOURVOS · Refer-DAVIS17 valJ: 41.23 -> 44.29Improved
fine-tune on Refer-DAVIS17 after pretrainingURVOS · Refer-DAVIS17 valJ: 44.29 -> 47.29Improved

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