Research map / Detection and segmentation / CNN features
Low-latency Event-based Object Detection with Spatially-Sparse Linear Attention
Haiqing Hao, Zhipeng Sui, Rong Zou, Zijia Dai, Nikola Zubic, Davide Scaramuzza, Wenhui Wang
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 spatially-sparse state activation (vs dense activation of all patches) (ablation) | SSLA-S · Gen1 val | MFLOPS/event: 388 -> 1.02 (no TD) | Improved |
| add spatially-sparse state activation (ablation) | SSLA-S · Gen1 val | mAP: 0.370 -> 0.370 | No clear effect |
| add temporal dropout (ablation) | SSLA-S · Gen1 val | MFLOPS/event: 1.02 -> 0.102 | Improved |
| add temporal dropout (ablation) | SSLA-S · Gen1 val | mAP: 0.370 -> 0.335 | Hurt |
| add sparse pooling (ablation) | SSLA-S · Gen1 val | mAP: 0.014 -> 0.335 | Improved |
| use SSLA instead of standard linear attention (ablation) | SSLA-S · Gen1 val | mAP: 0.001 -> 0.335 | Improved |
| add input position-aware projection (ablation) | SSLA-S · Gen1 val | mAP: 0.224 -> 0.335 | Improved |
| add output position-aware projection (ablation) | SSLA-S · Gen1 val | mAP: 0.306 -> 0.335 | Improved |
| increase patch size from 3 to 4 (ablation) | SSLA-S · Gen1 val | mAP: 0.335 -> 0.371 | Improved |
| decrease patch size from 3 to 2 (ablation) | SSLA-S · Gen1 val | mAP: 0.335 -> 0.200 | Hurt |
| replace LSTM with linear attention for parallel training (ablation) | SSLA-S · Gen1 | time per epoch (h): 1.05 -> 0.25 | Improved |
| replace LSTM with SSLA-S (ablation) | SSLA-S · Gen1 | mAP: 0.353 -> 0.334 | Hurt |
Closest papers by meaning
- SMamba: Sparse Mamba for Event-based Object Detection (AAAI 2025 · similarity 0.90)
- Scene Adaptive Sparse Transformer for Event-based Object Detection (CVPR 2024 · similarity 0.88)
- Efficient Event-Based Object Detection: A Hybrid Neural Network with Spatial and Temporal Attention (CVPR 2025 · similarity 0.88)
- FlexEvent: Towards Flexible Event-Frame Object Detection at Varying Operational Frequencies (NeurIPS 2025 · similarity 0.87)
- LEOD: Label-Efficient Object Detection for Event Cameras (CVPR 2024 · similarity 0.87)
- CREST: An Efficient Conjointly-trained Spike-driven Framework for Event-based Object Detection Exploiting Spatiotemporal Dynamics (AAAI 2025 · 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.
