Research map / Label-efficient and robust learning / Test-time adaptation
Context-Aware Streaming Perception in Dynamic Environments
Gur-Eyal Sela, Ionel Gog, Justin Wong, Kumar Krishna Agrawal, Xiangxi Mo, Sukrit Kalra, Peter Schafhalter, Eric Leong, Xin Wang, Bharathan Balaji, Joseph Gonzalez, Ion Stoica
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 |
|---|---|---|---|
| use Octopus context-based dynamic policy instead of global best static policy | streaming tracking pipeline · Waymo | S-MOTA: 25.1 -> 27.0 | Improved |
| use Octopus context-based dynamic policy instead of global best static policy | streaming tracking pipeline · Argoverse | S-MOTA: 49.4 -> 51.1 | Improved |
| use Octopus with ground truth features from current segment instead of prediction from previous segment (ablation) | Octopus · Waymo | S-MOTA: 27.0 -> 27.9 | Improved |
| use Octopus with ground truth features from current segment instead of prediction from previous segment (ablation) | Octopus · Argoverse | S-MOTA: 51.1 -> 53.2 | Improved |
| dynamically change the detection model only (ablation) | global best policy · Waymo | S-MOTA: 25.1 -> 27.6 | Improved |
| dynamically change tracking max age only (ablation) | global best policy · Waymo | S-MOTA: 25.1 -> 25.6 | Improved |
| dynamically change both detection model and tracking max age (ablation) | global best policy · Waymo | S-MOTA: 25.1 -> 27.9 | Improved |
| allow tracking maximum age to change in addition to detection model (ablation) | optimal dynamic policy · opportunity gap | S-MOTA: 28.3 -> 30.1 | Improved |
| allow tracking re-init frequency, min IOU and confidence threshold to change as well (ablation) | optimal dynamic policy · opportunity gap | S-MOTA: 30.1 -> 31.3 | Improved |
| remove baseline subtraction from regression targets (ablation) | Octopus regression · Waymo | S-MOTA: 27.9 -> 27.6 | Hurt |
| use independent classification instead of regression to rank configurations (ablation) | Octopus · Waymo | S-MOTA: 27.9 -> 27.2 | Hurt |
| use CNN based policy instead of engineered features with random forest (ablation) | policy model · Waymo | S-MOTA: 27.9 -> 25.2 | Hurt |
Closest papers by meaning
- Real-Time Object Detection for Streaming Perception (CVPR 2022 · similarity 0.89)
- Are We Ready for Vision-Centric Driving Streaming Perception? The ASAP Benchmark (CVPR 2023 · similarity 0.88)
- Real-time Stereo-based 3D Object Detection for Streaming Perception (NeurIPS 2024 · similarity 0.86)
- StreamForest: Efficient Online Video Understanding with Persistent Event Memory (NeurIPS 2025 · similarity 0.84)
- Transtreaming: Adaptive Delay-aware Transformer for Real-time Streaming Perception (AAAI 2025 · similarity 0.84)
- Glad: A Streaming Scene Generator for Autonomous Driving (ICLR 2025 · similarity 0.83)
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.
