An End-to-End Broad Learning System for Event-Based Object Classification

Abstract

Event cameras are bio-inspired vision sensors measuring brightness changes (referred to as an 'event') for each pixel independently, instead of capturing brightness images at a fixed rate using conventional cameras. Asynchronous event data mixed with noise information is challenging for event-based vision tasks. In this paper, we propose a broad learning network for object detection using the event data. The broad learning network consists of two distinct layers, a feature-node layer and an enhancement-node layer. Different to convolutional neural networks, the broad learning network can be extended by adding nodes into layers during training. We design a gradient descent algorithm to train network parameters, which creates an event-based broad learning network in an end-to-end manner. Our model outperforms state-of-the-art models, specifically, because of the small scale and increased speed displayed by our model during training. This demonstrates the superiority of event cameras towards online training and inference. INDEX TERMS Event camera, object classification, broad learning system, incremental learning.

BLCov: A novel collaborative–competitive broad learning system for COVID-19 detection from radiology images

“… Dang et al (2020) proposed a DWnet to recognize human actions, which feeds the extracted spatial–temporal features into BLS. BLS has also achieved good results in other tasks, such as event-based object classification ( Gao et al, 2020 ), Chinese herbal medicine classification ( Cai et al, 2019 ), and robotic material recognition ( Wang et al, 2019 ). BLS is also applied to many other fields, such as medical data analysis, system modeling, and fault detection ( Gong et al, 2019 , Chu et al, 2020 , Liu et al, 2019 ).…”

An Efficient Algorithm for the Incremental Broad Learning System by Inverse Cholesky Factorization of a Partitioned Matrix

“…When computed by (32a), the upper-triangular G can be computed by the inverse Choleksy factorization [37], or by inverting and transposing the lower-triangular Cholesky factor, as F i can be computed by inverting and transposing the Cholesky factor F −T i in (26). The matrix C has the same size as the added nodesH with only q columns, as shown in ( 16) and (33).…”

“…To solve the problem of small samples, a hidden layer has been added on the enhancement nodes of BLS in [25], for shallow learning of the enhanced features. For object detection using the event data, a BLS algorithm was proposed in [26], which includes a gradient descent algorithm to train network parameters. In [30], the defect of BLS was considered, i.e., the sensibility to the number of feature nodes, and then to dispose of this defect, a probabilistic model called sparse Bayesian BLS (SBBLS) was proposed to linearly-weight a small group of basis functions from an enormous number of candidates within Bayesian framework.…”

Fault Diagnosis of Sucker Rod Pump Based on Deep-Broad Learning Using Motor Data

“…Furthermore, the BLS extend the network structure using fast incremental learning without the retraining the full network. With the advantages of fast training and low rank approximation, the BLS has been widely used in pipeline leak detection [25], event camera [26] and fast PolSAR image classification [27] in last several years. BLS architecture: The outline of the entire BLS structure is illustrated in Fig.…”