Research map / Architectures and efficiency / RNNs and normalization
Hierarchical Neural Memory Network for Low Latency Event Processing
Ryuhei Hamaguchi, Yasutaka Furukawa, Masaki Onishi, Ken Sakurada
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 three stacked memories instead of one (ablation) | HMNet · DSEC-Semantic | accuracy: +3.8% | Improved |
| Use three stacked memories instead of one (ablation) | HMNet · DSEC-Semantic | accuracy: +3.2% | Improved |
| Use multi-level memory instead of single-level memory | HMNet · MVSEC outdoor day1 and outdoor night1 | depth estimation accuracy | Improved |
| Add the event gate to ESCA (ablation) | HMNet · DSEC-Semantic | accuracy: +1.6% | Improved |
| Add the event gate to ESCA (ablation) | HMNet · GEN1 | accuracy: +1.1% | Improved |
| Add down-write (ablation) | HMNet · DSEC-Semantic | accuracy: +1.6% | Improved |
| Add down-write (ablation) | HMNet · GEN1 | accuracy: +1.1% | Improved |
| Increase the z3 cycle length to 9 from shorter tested cycles (ablation) | HMNet-B3 · DSEC-Semantic and GEN1 | accuracy | Improved |
| Use a cycle length that is too long for z3 (ablation) | HMNet-B3 · GEN1 | accuracy | Hurt |
| Increase the inference time-step size from 5ms toward 6ms-8ms (ablation) | HMNet-B3/L3 · DSEC-Semantic and GEN1 | accuracy | Improved |
| Increase the inference time-step size beyond 10ms (ablation) | HMNet-B3/L3 · DSEC-Semantic and GEN1 | accuracy | Hurt |
| Fuse RGB frames by naive input concatenation | Baseline models · DSEC-Semantic | accuracy | Hurt |
Closest papers by meaning
- Efficient Learning of Event-based Dense Representation using Hierarchical Memories with Adaptive Update (ECCV 2024 · similarity 0.90)
- EventNet: Asynchronous Recursive Event Processing (CVPR 2019 · similarity 0.87)
- Leveraging Asynchronous Spiking Neural Networks for Ultra Efficient Event-Based Visual Processing (AAAI 2025 · similarity 0.86)
- Efficient Event-Based Object Detection: A Hybrid Neural Network with Spatial and Temporal Attention (CVPR 2025 · similarity 0.85)
- Efficient Event-Based Semantic Segmentation via Exploiting Frame-Event Fusion: A Hybrid Neural Network Approach (AAAI 2025 · similarity 0.84)
- MEET: Towards Memory-Efficient Temporal Sparse Deep Neural Networks (CVPR 2025 · similarity 0.84)
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.
