paper-with-me

Video Instance Segmentation 벤치마크

Video Instance Segmentation on OVIS validation

44개 결과 · ⬇ CSV · JSON

mask AP

14.3 25 35.7 46.4 57.1 2021-02 2026-11 CMaskTrack R-CNN (ResNet-50) — 15.4 (2021-02-02) CSipMask (ResNet-50) — 14.3 (2021-02-02) STMask(R101-DCN-FPN) — 17.3 (2021-04-06) CrossVIS (ResNet-50, calibration) — 18.1 (2021-04-13) CrossVIS (ResNet-50) — 14.9 (2021-04-13) D2Conv3D (ResNet-50) — 15.2 (2021-11-15) Mask2Former-VIS — 16.6 (2021-12-20) STC (ResNet-50) — 15.5 (2022-02-08) TeViT (ResNet-50) — 17.4 (2022-04-18) VITA (Swin-L) — 27.7 (2022-06-09) IDOL (Swin-L) — 42.6 (2022-07-21) IDOL (ResNet-50) — 30.2 (2022-07-21) DeVIS (Swin-L) — 35.5 (2022-07-22) DeVIS (ResNet-50) — 23.7 (2022-07-22) MinVIS (Swin-L) — 39.4 (2022-08-03) InstanceFormer (Swin-L) — 22.8 (2022-08-22) InstanceFormer(ResNet-50) — 20.0 (2022-08-22) GenVIS (Swin-L) — 45.4 (2022-11-16) ROVIS (Swin-L) — 42.6 (2022-11-16) TarViS (Swin-L) — 43.2 (2023-01-06) TarViS (Swin-T) — 34.0 (2023-01-06) TarViS (ResNet-50) — 31.1 (2023-01-06) UNINEXT (ViT-H, Online) — 49.0 (2023-03-12) UNINEXT (ResNet-50, Online) — 34.0 (2023-03-12) Tube-Link(ResNet-50) — 29.5 (2023-03-22) MDQE(SwinL) — 42.6 (2023-03-25) BoxVIS(Swin-L & Box-sup) — 40.6 (2023-03-26) GRAtt-VIS (Swin-L) — 45.7 (2023-05-26) GRAtt-VIS (ResNet-50) — 36.2 (2023-05-26) DVIS(Swin-L, Offline) — 49.9 (2023-06-06) DVIS(Swin-L, Online) — 47.1 (2023-06-06) RefineVIS (Swin-L, offline) — 46.0 (2023-06-07) CTVIS (Swin-L) — 46.9 (2023-07-24) CTVIS (ResNet-50) — 35.5 (2023-07-24) NOVIS (Swin-L) — 43.5 (2023-08-29) NOVIS (ResNet-50) — 32.7 (2023-08-29) GLEE-Pro — 50.4 (2023-12-14) DVIS++(VIT-L,Offline) — 53.4 (2023-12-20) DVIS++(VIT-L, Online) — 49.6 (2023-12-20) DVIS++(R50, Offline) — 41.2 (2023-12-20) DVIS++(R50, Online) — 37.2 (2023-12-20) UniVS(Swin-L) — 41.7 (2024-02-28) DVIS-DAQ(VIT-L, Offline) — 57.1 (2024-03-29) CAVIS(VIT-L, Offline) — 57.1 (2024-07-03) CMaskTrack R-CNN (ResNet-50) — 15.4 (2021-02-02) STMask(R101-DCN-FPN) — 17.3 (2021-04-06) CrossVIS (ResNet-50, calibration) — 18.1 (2021-04-13) VITA (Swin-L) — 27.7 (2022-06-09) IDOL (Swin-L) — 42.6 (2022-07-21) GenVIS (Swin-L) — 45.4 (2022-11-16) UNINEXT (ViT-H, Online) — 49.0 (2023-03-12) DVIS(Swin-L, Offline) — 49.9 (2023-06-06) GLEE-Pro — 50.4 (2023-12-14) DVIS++(VIT-L,Offline) — 53.4 (2023-12-20) DVIS-DAQ(VIT-L, Offline) — 57.1 (2024-03-29)
RankModel mask APAP50AP75APhoAPmoAR1APsoAR10 Extra Training Data PaperCodeYear
1 DVIS-DAQ(VIT-L, Offline) 57.183.862.9––––– ✓ DVIS-DAQ: Improving Video Segmentation via Dynamic Anchor Queries zhang-tao-whu/DVIS · zhang-tao-whu/DVIS_Plus · skyworkai/daq-vs 2024
1 CAVIS(VIT-L, Offline) 57.182.663.5––21.2–61.8 ✓ Context-Aware Video Instance Segmentation Seung-Hun-Lee/CAVIS 2024
3 DVIS++(VIT-L,Offline) 53.478.958.5––––– ✓ DVIS++: Improved Decoupled Framework for Universal Video Segmentation zhang-tao-whu/DVIS_Plus 2023
4 GLEE-Pro 50.4–55.5––––– ✓ General Object Foundation Model for Images and Videos at Scale FoundationVision/GLEE 2023
5 DVIS(Swin-L, Offline) 49.975.953.0––19.4–55.3 DVIS: Decoupled Video Instance Segmentation Framework zhang-tao-whu/DVIS 2023
6 DVIS++(VIT-L, Online) 49.672.555.027.156.620.869.954.6 ✓ DVIS++: Improved Decoupled Framework for Universal Video Segmentation zhang-tao-whu/DVIS_Plus 2023
7 UNINEXT (ViT-H, Online) 49.072.552.2––––– ✓ Universal Instance Perception as Object Discovery and Retrieval MasterBin-IIAU/UNINEXT 2023
8 DVIS(Swin-L, Online) 47.171.949.2––19.4–52.5 DVIS: Decoupled Video Instance Segmentation Framework zhang-tao-whu/DVIS 2023
9 CTVIS (Swin-L) 46.971.547.519.152.1––– ✓ CTVIS: Consistent Training for Online Video Instance Segmentation kainingying/ctvis 2023
10 RefineVIS (Swin-L, offline) 4670.448.4––19.1–51.2 ✓ RefineVIS: Video Instance Segmentation with Temporal Attention Refinement 2023
11 GRAtt-VIS (Swin-L) 45.769.147.8––19.2–49.4 ✓ GRAtt-VIS: Gated Residual Attention for Auto Rectifying Video Instance Segmentation tanveer81/grattvis 2023
12 GenVIS (Swin-L) 45.469.247.8––18.9–49.0 ✓ A Generalized Framework for Video Instance Segmentation miranheo/genvis 2022
13 NOVIS (Swin-L) 43.568.343.8––19.4–46.9 ✓ NOVIS: A Case for End-to-End Near-Online Video Instance Segmentation 2023
14 TarViS (Swin-L) 43.267.844.6––18.0–50.4 ✓ TarViS: A Unified Approach for Target-based Video Segmentation Ali2500/TarViS 2023
15 MDQE(SwinL) 42.667.844.321.649.318.365.146.5 MDQE: Mining Discriminative Query Embeddings to Segment Occluded Instances on Challenging Videos minghanli/mdqe_cvpr2023 2023
15 IDOL (Swin-L) 42.665.745.2––17.9–49.6 In Defense of Online Models for Video Instance Segmentation wjf5203/vnext · mkimhi/RISE 2022
15 ROVIS (Swin-L) 42.664.742.6––18.4–49.1 Robust Online Video Instance Segmentation with Track Queries zitongzhan/mmtracking 2022
18 UniVS(Swin-L) 41.7––––––– ✓ UniVS: Unified and Universal Video Segmentation with Prompts as Queries minghanli/univs 2024
19 DVIS++(R50, Offline) 41.268.940.9––16.8–47.3 ✓ DVIS++: Improved Decoupled Framework for Universal Video Segmentation zhang-tao-whu/DVIS_Plus 2023
20 BoxVIS(Swin-L & Box-sup) 40.668.439.920.945.8–59.4– BoxVIS: Video Instance Segmentation with Box Annotations minghanli/boxvis 2023
1–20 / 44 다음 → 페이지당 10 20 50 100