Research map / Video / Video segmentation and tracking
XMem++: Production-level Video Segmentation From Few Annotated Frames
Maksym Bekuzarov, Ariana Bermudez, Joon-Young Lee, Hao Li
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add permanent working memory storing annotated frames | XMem · LVOS validation, 10 annotated frames | J: 62.62 -> 75.74 | Improved |
| add permanent working memory storing annotated frames | XMem · MOSE, 5 annotated frames | J: 67.95 -> 77.11 | Improved |
| use XMem++ with 5 annotations instead of competitors with 10 | XMem++ · LVOS validation | J: 67.36 (5 frames) vs 65.07 best baseline (10 frames) | Improved |
| add permanent memory with one annotation | XMem · LVOS and MOSE, 1 frame | J | No clear effect |
| add permanent memory on challenging partial segmentation sequences, with fewer annotations | XMem++ · PUMaVOS | frame usage efficiency: 45 -> 10 annotations at similar J (85.43 vs 85.45) | Improved |
| use attention-based frame selection instead of experts | XMem++ · PUMaVOS user study | IoU: 0.805 (experts) vs 0.777 (algorithm) vs 0.779 (non-experts) | No clear effect |
| use attention-based frame selection instead of uniform sampling (ablation) | XMem++ · LVOS validation | J and F | No clear effect |
| disable temporary memory so only the annotated frames remain (ablation) | XMem++ · LVOS validation | segmentation quality | Improved |
| add Gaussian blur augmentation of annotated frames in permanent memory (ablation) | XMem++ · LVOS validation | segmentation quality: +0.019 | Improved |
| add translate right +100 px augmentation (ablation) | XMem++ · LVOS validation | segmentation quality: -0.037 | Hurt |
| add grayscale augmentation (ablation) | XMem++ · LVOS validation | segmentation quality: -0.005 | Hurt |
| add crop-mask-region augmentation (ablation) | XMem++ · LVOS validation | segmentation quality: -0.008 | Hurt |
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.
- XMem — XMem: Long-Term Video Object Segmentation with an Atkinson-Shiffrin Memory Model (ECCV 2022) · better on 8 of 9 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
- Video Object Segmentation without Temporal Information (TPAMI 2018 · similarity 0.90)
- XMem: Long-Term Video Object Segmentation with an Atkinson-Shiffrin Memory Model (ECCV 2022 · similarity 0.90)
- One-Shot Video Object Segmentation (CVPR 2017 · similarity 0.89)
- Space-Time Memory Networks for Video Object Segmentation With User Guidance (TPAMI 2020 · similarity 0.89)
- BubbleNets: Learning to Select the Guidance Frame in Video Object Segmentation by Deep Sorting Frames (CVPR 2019 · similarity 0.89)
- LiVOS: Light Video Object Segmentation with Gated Linear Matching (CVPR 2025 · similarity 0.88)
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.
