Research map / Video / Video segmentation and tracking
XMem: Long-Term Video Object Segmentation with an Atkinson-Shiffrin Memory Model
Ho Kei Cheng, Alexander G. Schwing
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 sensory memory (ablation) | XMem · Long-time Video (1x) | J&F: 87.9 -> 89.8 | Improved |
| add working memory (ablation) | XMem · Long-time Video (1x) | J&F: 38.7 -> 89.8 | Improved |
| add long-term memory (ablation) | XMem · DAVIS 2017 val | FPS: 17.6 -> 22.6 | Improved |
| add long-term memory (ablation) | XMem · YouTubeVOS 2018 val | G: 85.9 -> 85.7 | No clear effect |
| add both shrinkage and selection terms (ablation) | XMem · YouTubeVOS 2018 val | G: 85.0 -> 85.7 | Improved |
| add selection term only (ablation) | XMem · YouTubeVOS 2018 val | G: 85.0 -> 84.8 | Hurt |
| add shrinkage term only (ablation) | XMem · DAVIS 2017 val | J&F: 85.1 -> 85.6 | Improved |
| add potentiation to memory consolidation (ablation) | XMem · Long-time Video (3x) | J&F: 87.9 -> 90.0 | Improved |
| use usage-based instead of K-means prototype selection (P=256) (ablation) | XMem · Long-time Video (3x) | J&F: 74.5 -> 90.1 | Improved |
| use usage-based instead of random prototype selection (P=128) (ablation) | XMem · Long-time Video (3x) | J&F: 89.7 -> 90.0 | Improved |
| memory consolidation instead of eager compression (ablation) | XMem · Long-time Video (3x) | J&F: 87.3 -> 90.0 | Improved |
| memory consolidation instead of sparse insertion (ablation) | XMem · Long-time Video (3x) | J&F: 87.3 -> 90.0 | Improved |
Later papers that compared with this method
24 later accepted papers report XMem as a baseline in their results tables; 20 of them report a better number on most of the results they share with it.
- When Trackers Fail: VLM-Guided Verification and Recovery for Robust Video Object Segmentation (NeurIPS 2026 · better on 2 of 2 shared results)
- PanoSAM2: Lightweight Distortion- and Memory-aware Adaptions of SAM2 for 360 Video Object Segmentation (ECCV 2026 · better on 2 of 2 shared results)
- Advancing Complex Video Object Segmentation via Progressive Concept Construction (ICLR 2026 · better on 1 of 1 shared result)
- 360VOTS: Visual Object Tracking and Segmentation in Omnidirectional Videos (TPAMI 2025 · better on 5 of 5 shared results)
- GDKVM: Echocardiography Video Segmentation via Spatiotemporal Key-Value Memory with Gated Delta Rule (ICCV 2025 · better on 3 of 3 shared results)
- SMITE: Segment Me In TimE (ICLR 2025 · better on 1 of 2 shared results)
- SAM2Long: Enhancing SAM 2 for Long Video Segmentation with a Training-Free Memory Tree (ICCV 2025 · better on 2 of 2 shared results)
- RMem: Restricted Memory Banks Improve Video Object Segmentation (CVPR 2024 · better on 3 of 3 shared results)
16 more not listed.
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
- Learning Video Object Segmentation With Visual Memory (ICCV 2017 · similarity 0.91)
- XMem++: Production-level Video Segmentation From Few Annotated Frames (ICCV 2023 · similarity 0.90)
- Video Object Segmentation with Episodic Graph Memory Networks (ECCV 2020 · similarity 0.89)
- LiVOS: Light Video Object Segmentation with Gated Linear Matching (CVPR 2025 · similarity 0.89)
- Putting the Object Back into Video Object Segmentation (CVPR 2024 · similarity 0.88)
- RMem: Restricted Memory Banks Improve Video Object Segmentation (CVPR 2024 · similarity 0.87)
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.
