Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Dynamic Graph Induced Contour-aware Heat Conduction Network for Event-based Object Detection

Xiao Wang, Yu Jin, Lan Chen, Bo Jiang, Lin Zhu, Yonghong Tian, Jin Tang, Bin Luo

TL;DR

This work introduces CvHeat-DET, a contour-aware heat conduction network tailored for event-based object detection. It jointly learns dynamic contour graphs to guide a heat-diffusion backbone, predicts adaptive diffusivity via frequency embeddings, and fuses multi-scale contour and event representations through CHCO blocks. An IoU-based query selection scheme further enhances detector training. Across EvDET200K, GEN1, and DSEC, CvHeat-DET achieves state-of-the-art results on EvDET200K and demonstrates strong generalization, though online graph construction incurs slower inference. The approach highlights the value of contour-guided diffusion and multi-scale graph representations for robust event-based perception in real-world scenarios.

Abstract

Event-based Vision Sensors (EVS) have demonstrated significant advantages over traditional RGB frame-based cameras in low-light conditions, high-speed motion capture, and low latency. Consequently, object detection based on EVS has attracted increasing attention from researchers. Current event stream object detection algorithms are typically built upon Convolutional Neural Networks (CNNs) or Transformers, which either capture limited local features using convolutional filters or incur high computational costs due to the utilization of self-attention. Recently proposed vision heat conduction backbone networks have shown a good balance between efficiency and accuracy; however, these models are not specifically designed for event stream data. They exhibit weak capability in modeling object contour information and fail to exploit the benefits of multi-scale features. To address these issues, this paper proposes a novel dynamic graph induced contour-aware heat conduction network for event stream based object detection, termed CvHeat-DET. The proposed model effectively leverages the clear contour information inherent in event streams to predict the thermal diffusivity coefficients within the heat conduction model, and integrates hierarchical structural graph features to enhance feature learning across multiple scales. Extensive experiments on three benchmark datasets for event stream-based object detection fully validated the effectiveness of the proposed model. The source code of this paper will be released on https://github.com/Event-AHU/OpenEvDET.

Dynamic Graph Induced Contour-aware Heat Conduction Network for Event-based Object Detection

TL;DR

This work introduces CvHeat-DET, a contour-aware heat conduction network tailored for event-based object detection. It jointly learns dynamic contour graphs to guide a heat-diffusion backbone, predicts adaptive diffusivity via frequency embeddings, and fuses multi-scale contour and event representations through CHCO blocks. An IoU-based query selection scheme further enhances detector training. Across EvDET200K, GEN1, and DSEC, CvHeat-DET achieves state-of-the-art results on EvDET200K and demonstrates strong generalization, though online graph construction incurs slower inference. The approach highlights the value of contour-guided diffusion and multi-scale graph representations for robust event-based perception in real-world scenarios.

Abstract

Event-based Vision Sensors (EVS) have demonstrated significant advantages over traditional RGB frame-based cameras in low-light conditions, high-speed motion capture, and low latency. Consequently, object detection based on EVS has attracted increasing attention from researchers. Current event stream object detection algorithms are typically built upon Convolutional Neural Networks (CNNs) or Transformers, which either capture limited local features using convolutional filters or incur high computational costs due to the utilization of self-attention. Recently proposed vision heat conduction backbone networks have shown a good balance between efficiency and accuracy; however, these models are not specifically designed for event stream data. They exhibit weak capability in modeling object contour information and fail to exploit the benefits of multi-scale features. To address these issues, this paper proposes a novel dynamic graph induced contour-aware heat conduction network for event stream based object detection, termed CvHeat-DET. The proposed model effectively leverages the clear contour information inherent in event streams to predict the thermal diffusivity coefficients within the heat conduction model, and integrates hierarchical structural graph features to enhance feature learning across multiple scales. Extensive experiments on three benchmark datasets for event stream-based object detection fully validated the effectiveness of the proposed model. The source code of this paper will be released on https://github.com/Event-AHU/OpenEvDET.
Paper Structure (16 sections, 3 equations, 4 figures, 8 tables)

This paper contains 16 sections, 3 equations, 4 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Methodology
  4. Overview
  5. Input Encoding
  6. CvHeat-DET: Contour-aware Visual Heat Conduction for Event-based Detection
  7. IoU-based Query Selection
  8. Experiments
  9. Dataset and Evaluation Metric
  10. Implementation Details
  11. Comparison on Public Benchmarks
  12. Component Analysis
  13. Ablation Study
  14. Visualization
  15. Limitation Analysis
  16. ...and 1 more sections

Figures (4)

  • Figure 1: (a). Architectures of Self-Attention, Vanilla Heat Conduction Operator (HCO), and our proposed Contour-aware Heat Conduction Operator (HCO). (b). Information Conduction Mechanisms of Self-Attention, Vanilla Heat Conduction (HCO), and our proposed Contour-aware HCO.
  • Figure 2: An overview of our proposed event-based object detection framework, termed CvHeat-DET.
  • Figure 3: Visualization of the feature maps compared with other detectors.
  • Figure 4: Visualization of the detection results of ours and other detectors. (MC: misclassification, UD: undetected, OD: over-detected, LD: large deviation.)