IEEE Project Abstract

With the development of deep convolutional neural networks in recent years, the networkstructure has become more and more complicated and varied, and there are very good results in patternrecognition, image classification, scene classification and target tracking. This end-to-end learning modelrelies on the initial large dataset. However, many data are gradually obtained in practical situations, whichcontradicts the deep learning of one-time batch learning. There is an urgent need for an incrementallearning approach that can continuously learn new knowledge from new data while retaining what hasalready been learned. This paper proposes an incremental learning algorithm based on convolutional neuralnetwork and support vector data description. CNN and AM-Softmax loss function are used to represent andcontinuously learn image features. Support Vector Data Description (SVDD) is used to construct multiplehyperspheres for new and old classes of images. Class-incremental learning is achieved by the increment ofhyperspheres. The experimental results show that the incremental learning method proposed in this papercan effectively extract the latent features of the image and adapt it to the learning situation of the classincrement. The recognition accuracy is close to the batch learning.