Research map / Detection and segmentation / CNN features
Integer-Valued Training and Spike-driven Inference Spiking Neural Network for High-performance and Energy-efficient Object Detection
Xinhao Luo, Man Yao, Yuhong Chou, Bo Xu, Guoqi Li
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 SpikeYOLO architecture instead of directly spiking-converted YOLOv8 (ablation) | spiking YOLO on COCO · COCO 2017 val | mAP@50 (%): 46.8 -> 52.7 | Improved |
| expand integer value D from 1 to 4 (I-LIF) (ablation) | SpikeYOLO 23.1M, T=1 · COCO 2017 val | mAP@50: +9.6% | Improved |
| increase timesteps T from 1 to 4 with D=1 (ablation) | SpikeYOLO 23.1M · COCO 2017 val | mAP@50: 52.7 -> 55.7 (+3%) | Improved |
| increase D instead of T for added capacity (ablation) | SpikeYOLO · COCO 2017 val | power: 1x1 to 1x4: +88% power; 1x1 to 4x1: +267% power | Improved |
| increase D at fixed T=2 (ablation) | SpikeYOLO 23.1M · Gen1 | mAP@50 and power: 63.6 -> 66.1 -> 67.0 mAP@50; power 8.1 -> 7.8 -> 7.1 mJ | Improved |
| increase T from 4 to 8 (ablation) | SpikeYOLO 23.1M, D=1 · Gen1 | mAP@50: +0.7% | No clear effect |
| increase T from 1 to 4 (ablation) | SpikeYOLO 23.1M, D=1 · Gen1 | mAP@50: +6.7% | Improved |
| add re-parameterization convolution (ablation) | SpikeYOLO · COCO 2017 val | mAP@50: 60.6 -> 62.3 | Improved |
| use standard-convolution SNN-Block-1 in low stages (ablation) | SpikeYOLO · COCO 2017 val | mAP@50: 61.2 -> 62.3 | Improved |
| replace high-stage SNN-Block-2 with transformer-based SNN block (ablation) | SpikeYOLO · COCO 2017 val | mAP@50: 62.3 -> 61.0 | Hurt |
| use anchor-free head instead of anchor-based head (ablation) | SpikeYOLO · COCO 2017 val | mAP@50:95: 39.7 -> 45.5 | Improved |
| run SpikeYOLO as SNN instead of ANN of same architecture | SpikeYOLO 23.1M · Gen1 | mAP@50: 64.7 -> 67.2 | Improved |
Closest papers by meaning
- Brain-Inspired Spiking Neural Networks for Energy-Efficient Object Detection (CVPR 2025 · similarity 0.94)
- SpikingYOLOX: Improved YOLOX Object Detection with Fast Fourier Convolution and Spiking Neural Networks (AAAI 2025 · similarity 0.93)
- SFOD: Spiking Fusion Object Detector (CVPR 2024 · similarity 0.92)
- Efficient Event-Based Object Detection: A Hybrid Neural Network with Spatial and Temporal Attention (CVPR 2025 · similarity 0.90)
- Inference-Scale Complexity in ANN-SNN Conversion for High-Performance and Low-Power Applications (CVPR 2025 · similarity 0.90)
- CREST: An Efficient Conjointly-trained Spike-driven Framework for Event-based Object Detection Exploiting Spatiotemporal Dynamics (AAAI 2025 · similarity 0.90)
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.
