Research map / Probabilistic and scientific modeling / Dynamical systems
Continuous Spatiotemporal Events Decoupling through Spike-based Bayesian Computation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Builds a spiking neural network that performs Bayesian EM-style motion segmentation of event-camera streams, with WTA circuits as the E-step and STDP as the M-step for contrast maximization.
Problem
Decoupling and segmenting continuous spatiotemporal event streams generated by different motions (camera ego-motion and moving objects) without prior knowledge, using a spiking Bayesian computation model.
Approach
Maps the EM-based contrast-maximization motion segmentation onto a spiking network: WTA circuits with global inhibition compute per-event responsibilities (E-step), and STDP updates motion parameters so that only events matching a motion pattern drive its update (M-step), with theoretical argument that the STDP update increases IWE variance.
Key results
- STDP-based learning increases the variance (contrast) of warped event images, shown analytically via a first-order approximation.
- On three EED scenes with mixed camera self-motion and moving objects, output neurons separate moving-object motion parameters online.
- The paper explicitly states limited quantitative evaluation and no comparison with existing motion segmentation algorithms.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use the spike-based Bayesian computation network for online motion segmentation | Event streams in EED · Extreme Event Dataset (EED) | motion segmentation | Helped |
| Replace SVD with K-means for selecting initial motion parameters (ablation) | Spike-based Bayesian Computation · EED parameter initialization | similarity of selected parameter points | No clear effect |
Limitations
- Limited quantitative evaluation and no comparison with existing event-based motion segmentation methods.
- Local learning rules may not converge to global optimum; highly dependent on parameter initialization.
- Restricted to linear motion models in experiments.
Experimental setup
Datasets: Extreme Event Dataset (EED)
Benchmarks: Extreme Event Dataset (EED)
Metrics: qualitative segmentation examples IWE contrast/variance
Builds on: Expectation-Maximization contrast maximization spike-based Bayesian computation (Nessel et al.) STDP
Closest papers by meaning
- A Scalable, Causal, and Energy Efficient Framework for Neural Decoding with Spiking Neural Networks (NeurIPS 2025 · similarity 0.88)
- Spiking Neural Network as Adaptive Event Stream Slicer (NeurIPS 2024 · similarity 0.86)
- Leveraging Asynchronous Spiking Neural Networks for Ultra Efficient Event-Based Visual Processing (AAAI 2025 · similarity 0.86)
- A Chaotic Dynamics Framework Inspired by Dorsal Stream for Event Signal Processing (ICML 2025 · similarity 0.86)
- SFOD: Spiking Fusion Object Detector (CVPR 2024 · similarity 0.86)
- Spatio-Temporal Interactive Learning for Efficient Image Reconstruction of Spiking Cameras (NeurIPS 2024 · similarity 0.86)
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.