Spiking Neural Network as Adaptive Event Stream Slicer

Authors: Jiahang Cao, Mingyuan Sun, Ziqing Wang, Hao Cheng, Qiang Zhang, shibo zhou, Renjing Xu

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments demonstrate that our method yields significant performance improvements in event-based object tracking and recognition.
Researcher Affiliation Collaboration Jiahang Cao1 Mingyuan Sun2 Ziqing Wang3 Hao Cheng1 Qiang Zhang1,4 Shibo Zhou5 Renjing Xu1 1 The Hong Kong University of Science and Technology (Guangzhou) 2 Northeastern University 3 Northwestern University 4 Beijing Innovation Center of Humanoid Robotics Co. Ltd. 5 Brain Mind Innovation
Pseudocode Yes Algorithm 1 Adaptive Event Slicing Process and Algorithm 2 Feedback-Update Training Strategy
Open Source Code Yes Our code is available at https://github.com/Andy Cao1125/Spike Slicer.
Open Datasets Yes Datasets. The FE108 dataset [3] is an extensive event-based dataset for single object tracking... The DVS-Gesture [34] dataset contains 11 hand gestures... The N-Caltech101 dataset [35] incorporates 8,831 event-based images... The DVS-CIFAR10 dataset [36] is an event-stream dataset... The SL-Animal database [37] features DVS recordings...
Dataset Splits No We choose 54 sequences for training ANNs, 22 sequences for training SNNs and the rest 32 sequences for testing.
Hardware Specification Yes Each experiment is conducted in an NVIDIA 4090 GPU.
Software Dependencies No No specific software dependencies with version numbers (e.g., PyTorch, Python versions) are mentioned.
Experiment Setup Yes We adopt the SGD optimizer and set the initial learning rate as 1e-4, along with the cosine learning rate scheduler. SNN models are trained with batch size 32.