Papers › The Devil is in Temporal Token: High Quality Video Reasoning Segmentation

The Devil is in Temporal Token: High Quality Video Reasoning Segmentation

15 Jan 2025CVPR 2025 1arXiv:2501.08549archive 2025-07-28

Sitong Gong, Yunzhi Zhuge, Lu Zhang, Zongxin Yang, Pingping Zhang, Huchuan Lu

Existing methods for Video Reasoning Segmentation rely heavily on a single special token to represent the object in the keyframe or the entire video, inadequately capturing spatial complexity and inter-frame motion. To overcome these challenges, we propose VRS-HQ, an end-to-end video reasoning segmentation approach that leverages Multimodal Large Language Models (MLLMs) to inject rich spatiotemporal features into hierarchical tokens.Our key innovations include a Temporal Dynamic Aggregation (TDA) and a Token-driven Keyframe Selection (TKS). Specifically, we design frame-level <SEG> and temporal-level <TAK> tokens that utilize MLLM's autoregressive learning to effectively capture both local and global information. Subsequently, we apply a similarity-based weighted fusion and frame selection strategy, then utilize SAM2 to perform keyframe segmentation and propagation. To enhance keyframe localization accuracy, the TKS filters keyframes based on SAM2's occlusion scores during inference. VRS-HQ achieves state-of-the-art performance on ReVOS, surpassing VISA by 5.9%/12.5%/9.1% in J&F scores across the three subsets. These results highlight the strong temporal reasoning and segmentation capabilities of our method. Code and model weights will be released at VRS-HQ.

PaperPDFConference PDFCode

In Syntology View this paper on Syntology: its repositories, every harvested function with whether it ran, its licence and the call to fetch it.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

sitonggong/vrs-hq officialmentioned in papermentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Reasoning SegmentationReferring Expression SegmentationReferring Video Object SegmentationSegmentation

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Referring Expression Segmentation Refer-YouTube-VOS (2021 public validation) VRS-HQ (Chat-UniVi-13B) F 73.1 #2 of 33 Archive leaderboard report
Referring Expression Segmentation Refer-YouTube-VOS (2021 public validation) VRS-HQ (Chat-UniVi-13B) J 69 #2 of 33 Archive leaderboard report
Referring Expression Segmentation Refer-YouTube-VOS (2021 public validation) VRS-HQ (Chat-UniVi-13B) J&F 71 #2 of 33 Archive leaderboard report
Referring Video Object Segmentation MeViS VRS-HQ (Chat-UniVi-13B) F 53.7 #4 of 16 Archive leaderboard report
Referring Video Object Segmentation MeViS VRS-HQ (Chat-UniVi-13B) J 48 #4 of 16 Archive leaderboard report
Referring Video Object Segmentation MeViS VRS-HQ (Chat-UniVi-13B) J&F 50.9 #4 of 16 Archive leaderboard report
Referring Video Object Segmentation ReVOS VRS-HQ (Chat-UniVi-13B) F 62.5 #1 of 9 Archive leaderboard report
Referring Video Object Segmentation ReVOS VRS-HQ (Chat-UniVi-13B) J 57.6 #1 of 9 Archive leaderboard report
Referring Video Object Segmentation ReVOS VRS-HQ (Chat-UniVi-13B) J&F 60 #1 of 9 Archive leaderboard report
Referring Video Object Segmentation ReVOS VRS-HQ (Chat-UniVi-13B) R 18.9 #1 of 9 Archive leaderboard report
Referring Video Object Segmentation ReVOS VRS-HQ (Chat-UniVi-7B) F 61.6 #2 of 9 Archive leaderboard report
Referring Video Object Segmentation ReVOS VRS-HQ (Chat-UniVi-7B) J 56.6 #2 of 9 Archive leaderboard report
Referring Video Object Segmentation ReVOS VRS-HQ (Chat-UniVi-7B) J&F 59.1 #2 of 9 Archive leaderboard report
Referring Video Object Segmentation ReVOS VRS-HQ (Chat-UniVi-7B) R 19.7 #2 of 9 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections