Research map / Detection and segmentation / CNN features
Rethinking Neuromorphic Object Detection with Hybrid Dynamic Interaction Transformers
Dianze Li, Jianing Li, Xu Liu, Zhaokun Zhou, Xiaopeng Fan B, Yonghong Tian
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 relative semantic embedding (semantic-enhanced self-attention) (ablation) | Swin Transformer baseline · DSEC-Detection | mAP: 0.269 -> 0.277 | Improved |
| add event stream through the Spiking Swin Transformer branch (ablation) | HDI-Former · DSEC-Detection | mAP: 0.277 -> 0.284 | Improved |
| add dynamic interaction between ANN and SNN branches (ablation) | HDI-Former · DSEC-Detection | mAP: 0.284 -> 0.291 | Improved |
| add dynamic interaction between ANN and SNN branches (ablation) | HDI-Former · DSEC-Detection | mAP50: 0.511 -> 0.522 | Improved |
| add dynamic interaction between ANN and SNN branches (ablation) | HDI-Former · PKU-DAVIS-SOD | mAP50: 0.501 -> 0.516 | Improved |
| use Spiking Swin Transformer in place of EMS-YOLO for event input (ablation) | SNN branch · DSEC-Detection | mAP: 0.082 -> 0.116 | Improved |
| use spiking instead of ANN Swin-T on event data (ablation) | Swin-T · Gen1 | mAP: 0.296 -> 0.291 | Hurt |
| use spiking instead of ANN Swin-T on event data (ablation) | Swin-T · DSEC-Detection (events) | mAP: 0.133 -> 0.116 | Hurt |
| set SNN timestep T=5 instead of 1 (ablation) | Spiking Swin-T · Gen1 | mAP: 0.286 -> 0.291 | Improved |
| set SNN timestep T=5 instead of 9 (ablation) | Spiking Swin-T · Gen1 | mAP: 0.286 -> 0.291 | Improved |
| apply dynamic interaction in 4 layers instead of 0 (ablation) | HDI-Former · DSEC-Detection | mAP: 0.284 -> 0.291 | Improved |
| apply dynamic interaction in 8 layers instead of 4 (ablation) | HDI-Former · DSEC-Detection | mAP: 0.291 -> 0.286 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 4 on most shared results.
- SODFormer — SODFormer: Streaming Object Detection With Transformer Using Events and Frames (TPAMI 2023) · better on 3 of 3 shared results
- SFNet — SFNet: Learning Object-Aware Semantic Correspondence (CVPR 2019) · better on 2 of 2 shared results
- YOLOv7 — YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors (CVPR 2023) · better on 1 of 1 shared result
- Faster R-CNN — Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (TPAMI 2016) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Efficient Event-Based Object Detection: A Hybrid Neural Network with Spatial and Temporal Attention (CVPR 2025 · similarity 0.91)
- Hybrid Spiking Vision Transformer for Object Detection with Event Cameras (ICML 2025 · similarity 0.88)
- SFOD: Spiking Fusion Object Detector (CVPR 2024 · similarity 0.88)
- Brain-Inspired Spiking Neural Networks for Energy-Efficient Object Detection (CVPR 2025 · similarity 0.86)
- CREST: An Efficient Conjointly-trained Spike-driven Framework for Event-based Object Detection Exploiting Spatiotemporal Dynamics (AAAI 2025 · similarity 0.86)
- Shrinking Your TimeStep: Towards Low-Latency Neuromorphic Object Recognition with Spiking Neural Networks (AAAI 2024 · similarity 0.85)
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.
