Research map / Detection and segmentation / CNN features
Real-Time Object Detection for Streaming Perception
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 |
|---|---|---|---|
| Directly predict next-frame detections rather than predict current boxes plus offsets (ablation) | YOLOX-L · Argoverse-HD | sAP: 31.0 -> 34.2 | Helped |
| Fuse adjacent-frame features at the FPN level rather than at input or backbone (ablation) | YOLOX-L · Argoverse-HD | sAP: 30.3 -> 34.2; 30.5 -> 34.2 | Helped |
| Replace FPN element-wise addition with concatenation (ablation) | YOLOX-L · Argoverse-HD | sAP: 30.8 -> 34.2 | Helped |
| Add DFP to the future-prediction pipeline (ablation) | YOLOX-L · Argoverse-HD | sAP: 34.2 -> 35.5 | Helped |
| Add TAL to the future-prediction pipeline (ablation) | YOLOX-L · Argoverse-HD | sAP: 34.2 -> 35.1 | Helped |
| Add both DFP and TAL to the future-prediction pipeline (ablation) | YOLOX-L · Argoverse-HD | sAP: 34.2 -> 36.1 | Helped |
| Add DFP to the pipeline for static-scene prediction (ablation) | YOLOX-L · Argoverse-HD, 0x speed | sAP: 36.4 -> 38.3 | Helped |
| Add DFP and TAL to the pipeline at faster driving speed (ablation) | YOLOX-L · Argoverse-HD, 2x speed | sAP: 31.3 -> 33.3 | Helped |
| Increase attention to new objects by setting ν below 1 (ablation) | TAL · Argoverse-HD | sAP: 36.1 -> 35.0 | Hurt |
| Use end-to-end forecasting instead of Kalman-filter forecasting | YOLOX-L streaming detector · Argoverse-HD, 1x speed | sAP: 35.5 -> 36.1 | Helped |
| Use end-to-end forecasting instead of Kalman-filter forecasting | YOLOX-L streaming detector · Argoverse-HD, 2x speed | sAP: 31.8 -> 33.3 | Helped |
| Use end-to-end forecasting instead of Kalman-filter forecasting | YOLOX-L streaming detector · Argoverse-HD | Extra Latency: 3.11 ms -> 0.8 ms | Helped |
Closest papers by meaning
- Real-time Stereo-based 3D Object Detection for Streaming Perception (NeurIPS 2024 · similarity 0.90)
- Transtreaming: Adaptive Delay-aware Transformer for Real-time Streaming Perception (AAAI 2025 · similarity 0.90)
- StreamForest: Efficient Online Video Understanding with Persistent Event Memory (NeurIPS 2025 · similarity 0.86)
- Glad: A Streaming Scene Generator for Autonomous Driving (ICLR 2025 · similarity 0.83)
- VISTREAM: Improving Computation Efficiency of Visual Streaming Perception via Law-of-Charge-Conservation Inspired Spiking Neural Network (CVPR 2025 · similarity 0.83)
- StreamFlow: Streamlined Multi-Frame Optical Flow Estimation for Video Sequences (NeurIPS 2024 · 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.