Image Recognition with Histogram of Oriented Gradient Feature and Pseudoinverse Learning AutoEncoders

Neural network is an artificial intelligence technology which achieve good results in computer vision, natural language processing and other related fields. Currently the most used model for image recognition is convolutional neural networks, however, it has complex structure, there many group open sources of code but it is difficult to reuse. Moreover, most of training algorithm of the model is based on the gradient descent which takes a lot of time to adjust parameters. In order to solve these problems, this paper presents a model combining the histogram of oriented gradient and the pseudoinverse learning autoencoders. Our model does not require any iterative optimization, the number of the neurons and the number of hidden layers are automatically determined in the model. At the same time, our model has a simple structure, do not requires a huge amount of computing resources. Experimental results show that our model is superior to other baseline models.