Asynchronous Convolutional Networks for Object Detection in Neuromorphic Cameras

TL;DR

This paper introduces two neural network architectures for object detection in neuromorphic cameras, leveraging event-based data and sparsity to improve detection efficiency and accuracy.

Contribution

It proposes YOLE and fcYOLE architectures, with fcYOLE being an asynchronous, fully convolutional network that exploits event sparsity for better performance.

Findings

fcYOLE outperforms frame-based models in detection accuracy

The models effectively utilize event sparsity for computational efficiency

Evaluation on multiple datasets demonstrates robustness and versatility

Abstract

Event-based cameras, also known as neuromorphic cameras, are bioinspired sensors able to perceive changes in the scene at high frequency with low power consumption. Becoming available only very recently, a limited amount of work addresses object detection on these devices. In this paper we propose two neural networks architectures for object detection: YOLE, which integrates the events into surfaces and uses a frame-based model to process them, and fcYOLE, an asynchronous event-based fully convolutional network which uses a novel and general formalization of the convolutional and max pooling layers to exploit the sparsity of camera events. We evaluate the algorithm with different extensions of publicly available datasets and on a novel synthetic dataset.