CIFAR10-DVS: An Event-Stream Dataset for Object Classification

Abstract

Neuromorphic vision research requires high-quality and appropriately challenging event-stream datasets to support continuous improvement of algorithms and methods. However, creating event-stream datasets is a time-consuming task, which needs to be recorded using the neuromorphic cameras. Currently, there are limited event-stream datasets available. In this work, by utilizing the popular computer vision dataset CIFAR-10, we converted 10,000 frame-based images into 10,000 event streams using a dynamic vision sensor (DVS), providing an event-stream dataset of intermediate difficulty in 10 different classes, named as “CIFAR10-DVS.” The conversion of event-stream dataset was implemented by a repeated closed-loop smooth (RCLS) movement of frame-based images. Unlike the conversion of frame-based images by moving the camera, the image movement is more realistic in respect of its practical applications. The repeated closed-loop image movement generates rich local intensity changes in continuous time which are quantized by each pixel of the DVS camera to generate events. Furthermore, a performance benchmark in event-driven object classification is provided based on state-of-the-art classification algorithms. This work provides a large event-stream dataset and an initial benchmark for comparison, which may boost algorithm developments in even-driven pattern recognition and object classification.

Asynchronous Convolutional Networks for Object Detection in Neuromorphic Cameras

“…Only few event-based object recognition datasets are publicly available in the literature. The most popular ones are: N-MNIST [5], MNIST-DVS [6], CIFAR10-DVS [17] and N-Caltech101 [5]. These datasets are obtained from the original MNIST [3], CIFAR-10 [12] and Cal-tech101 [6] datasets by recording the original images with an event camera while moving the camera itself or the images of the datasets.…”

Condition-Guided Adversarial Generative Testing for Deep Learning Systems

“…For research purpose, we adopt two popular and commonly used datasets with different types of data: MNIST [21] and CIFAR-10 [22]. At the same time, we have learned and trained several popular DNN models for each dataset, which have been widely used by scientific researchers.…”

“…We construct three different kinds of neural networks based on LeNet family, namely LeNet-1, LeNet-4 and LeNet-5. CIFAR-10 [22] is a set of general image classification images, including 32 * 32 * 3 pixel three-channel images, including ten different kinds of pictures (such as aircraft, cats, trucks, etc.). The dataset contains 50,000 training examples and 10,000 test examples.…”

“…The coverage-guided adversarial generative fuzzy testing framework is provided in Section 3. In Section 4, we use two popular datasets (MNIST [21] and Cifar-10 [22]) to validate our approach. Existing work and their limitations are discussed in Section 5.…”

Applications of event‐based image sensors—Review and analysis

“…Images generated with asynchronous image sensors are ideal for this purpose. The spread and popularization of DVSs among the community have driven the creation of event‐based image datasets recorded with DVSs Traffic monitoring.…”