Small-scale image recognition based on Cascaded Convolutional Neural Network

The advent of the era of big data and the rapid improvement of computer computing capabilities have promoted the development of image recognition technology in a more advanced direction. Image recognition technology based on deep learning has become a current research hotspot in the field of artificial intelligence. As one of the algorithms of deep learning, convolutional neural network has been widely used in the field of image recognition due to its superior performance. This paper designs a convolutional neural network that can take into account model size and accuracy. It is suitable for small-scale image recognition and is convenient for deployment in some environments with low hardware platforms. Based on the improvement of the classic ALexNet network, a parallel convolutional structure is used to design a parallel cascaded convolutional network based on jumper connections for small-scale image recognition. Parallel convolution uses convolution kernels of different scales to extract features in parallel, and cascade fusion the features extracted from different scales. In order to further improve the accuracy, the network was optimized by adding a layer connection, and the network performance was evaluated on the Caltech-256 and Food-101 data sets. The results show that compared with the classic AlexNet network, the network RPCNet constructed based on the layer connection has improved accuracy by 6.12% and 12.28%, respectively, and the network scale is only 1/15 of the ALexNet network.