Research map / Video / Video segmentation and tracking
End-to-End Video Instance Segmentation With Transformers
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use ResNet-101 instead of ResNet-50 | VisTR · YouTube-VIS val | AP: 36.2 -> 40.1 | Helped |
| Increase the input sequence length from 18 to 36 frames (ablation) | VisTR · YouTube-VIS val | AP: 29.7 -> 33.3 | Helped |
| Share instance-level queries across frames instead of using one query per prediction (ablation) | VisTR · YouTube-VIS val | AP: 33.3 -> 32.0 | Hurt |
| Share one query embedding across all predictions (ablation) | VisTR · YouTube-VIS val | AP: 33.3 -> 8.4 | Hurt |
| Share query embeddings by frame (ablation) | VisTR · YouTube-VIS val | AP: 33.3 -> 13.7 | Hurt |
| Randomize the temporal order of input frames (ablation) | VisTR · YouTube-VIS val | AP: 32.3 -> 33.3 | Hurt |
| Remove spatial and temporal positional encoding (ablation) | VisTR · YouTube-VIS val | AP: 28.4 -> 33.3 | Hurt |
| Use CNN-encoded rather than Transformer-encoded features for mask prediction (ablation) | VisTR · YouTube-VIS val | AP: 32.0 -> 33.3 | Hurt |
| Replace 3D instance-sequence segmentation with per-frame 2D convolution (ablation) | VisTR · YouTube-VIS val | AP: 33.3 -> 34.4 | Hurt |
| Use VisTR instead of MaskTrack R-CNN with ResNet-50 | Video instance segmentation · YouTube-VIS val | AP: 30.3 -> 36.2 | Helped |
Closest papers by meaning
- Efficient Video Instance Segmentation via Tracklet Query and Proposal (CVPR 2022 · similarity 0.94)
- Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation (CVPR 2025 · similarity 0.89)
- SyncVIS: Synchronized Video Instance Segmentation (NeurIPS 2024 · similarity 0.88)
- Decoupled Motion Expression Video Segmentation (CVPR 2025 · similarity 0.88)
- Robust and Consistent Online Video Instance Segmentation via Instance Mask Propagation (AAAI 2025 · similarity 0.88)
- VideoCutLER: Surprisingly Simple Unsupervised Video Instance Segmentation (CVPR 2024 · 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.