Research map / Video / Video-language understanding
Towards Streaming Perception
Mengtian Li, Yu-Xiong Wang, Deva Ramanan
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 |
|---|---|---|---|
| move from offline to streaming evaluation | streaming detector · Argoverse-HD | AP: 38.0 -> 6.2 | Hurt |
| add scheduling, association and forecasting | streaming detector · Argoverse-HD | AP: 6.2 -> 17.8 | Improved |
| add infinite GPUs | optimized streaming pipeline · Argoverse-HD | AP: 17.8 -> 20.3 | Improved |
| add dynamic shrinking-tail scheduling | Mask R-CNN R50 @ s0.5 · Argoverse-HD instance segmentation | AP: 17.2 -> 18.3 | Improved |
| add infinite GPUs | Mask R-CNN R50 instance segmentation · Argoverse-HD instance segmentation | AP: 18.3 -> 20.6 | Improved |
| add dynamic scheduling | detector on Tesla V100 · Argoverse-HD V100 | AP: 13.0 -> 14.0 | Improved |
| add forecasting to visual tracking (ablation) | detection + visual tracking · Argoverse-HD V100 | AP: 12.6 -> 18.0 | Improved |
| add forecasting to visual tracking with infinite GPUs (ablation) | visual tracking · Argoverse-HD V100 | AP: 14.4 -> 22.8 | Improved |
| add Kalman filter forecasting (ablation) | Mask R-CNN R50 @ s0.5 with scheduling · Argoverse-HD streaming | AP: 13.0 -> 16.7 | Improved |
| use linear instead of Kalman forecasting (ablation) | streaming forecasting · Argoverse-HD streaming | AP: 16.7 -> 15.7 | Hurt |
| use quadratic extrapolation forecasting (ablation) | streaming forecasting · Argoverse-HD streaming | AP: 13.0 -> 9.7 | Hurt |
| use end-to-end F2F instead of Kalman filter (ablation) | offline forecasting · Argoverse-HD offline | AP: 19.1 -> 18.3 | Hurt |
Closest papers by meaning
- Real-time Stereo-based 3D Object Detection for Streaming Perception (NeurIPS 2024 · similarity 0.80)
- Don't Pause! Every prediction matters in a streaming video (NeurIPS 2026 · similarity 0.80)
- StreamForest: Efficient Online Video Understanding with Persistent Event Memory (NeurIPS 2025 · similarity 0.80)
- StreamOV: Streaming Omni-Video Understanding via Evidence-Guided Memory and Response Triggering (NeurIPS 2026 · similarity 0.79)
- What Should a Streaming Video Model Remember? (NeurIPS 2026 · similarity 0.78)
- StreamGaze: Gaze-Guided Temporal Reasoning and Proactive Understanding in Streaming Videos (NeurIPS 2026 · similarity 0.78)
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.
