Research map / Video / Video segmentation and tracking
S2N: Suppression-Strengthen Network for Event-Based Recognition under Variant Illuminations
Zengyu Wan, Yang Wang, Ganchao Tan, Yang Cao, Zheng-Jun Zha
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 exposure loss Le to NSN training (ablation) | S2N · DVSGesture ablation (train S4, test S2 natural) | accuracy (%): 67.0 -> 92.1 | Improved |
| add Ln loss to NSN training (ablation) | S2N · DVSGesture ablation (train S4, test S2 natural) | accuracy (%): 86.6 -> 92.1 | Improved |
| add Lc loss to NSN training (ablation) | S2N · DVSGesture ablation (train S4, test S2 natural) | accuracy (%): 88.2 -> 92.1 | Improved |
| add Noise Suppression Network in front of FSN (ablation) | S2N · DVSGesture ablation (train S4, test S2 natural) | accuracy (%): 84.1 -> 92.1 | Improved |
| replace I3D with Feature Strengthen Network after NSN (ablation) | S2N · DVSGesture ablation (train S4, test S2 natural) | accuracy (%): 75.5 -> 92.1 | Improved |
| add Evolution-aware Module to FSN (ablation) | FSN · DVSGesture ablation (train S4, test S2 natural) | accuracy (%): 77.4 -> 84.1 | Improved |
| use density-adaptive CDC instead of CDC (ablation) | FSN · DVSGesture ablation (train S4, test S2 natural) | accuracy (%): 74.5 -> 77.4 | Improved |
| add NSN denoising to I3D | I3D · DVSGesture (train S4, test S3) | accuracy (%): 78.4 -> 82.0 | Improved |
| add VN denoising to I3D | I3D · DVSGesture (train S4, test S3) | accuracy (%): 78.4 -> 76.2 | Hurt |
| add YN denoising to I3D | I3D · DVSGesture (train S4, test S0) | accuracy (%): 90.3 -> 83.7 | Hurt |
| add VN denoising to EST | EST · DAVIS346Character (train L0 300lux, test all) | accuracy (%): 29.4 -> 47.7 | Improved |
| replace denoise+recognizer pipelines with S2N | S2N · DVSGesture (train S4 lab, test all 5 scenes) | accuracy (%): 85.3 -> 94.7 | Improved |
Closest papers by meaning
- Event Enhanced High-Quality Image Recovery (ECCV 2020 · similarity 0.85)
- EventRPG: Event Data Augmentation with Relevance Propagation Guidance (ICLR 2024 · similarity 0.84)
- Towards Robust Event-guided Low-Light Image Enhancement: A Large-Scale Real-World Event-Image Dataset and Novel Approach (CVPR 2024 · similarity 0.83)
- LED: A Large-scale Real-world Paired Dataset for Event Camera Denoising (CVPR 2024 · similarity 0.83)
- SpikePoint: An Efficient Point-based Spiking Neural Network for Event Cameras Action Recognition (ICLR 2024 · similarity 0.83)
- Efficient Event-Based Object Detection: A Hybrid Neural Network with Spatial and Temporal Attention (CVPR 2025 · similarity 0.82)
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.
