This project is based on Pytorch to study the image classification algorithm and realizes the image classification and the recognition of Intel Image Classification by using nn to build a fully connected neural network model. As an alternative to Numpy, Pytorch has advanced features and supports GPU acceleration, so it can quickly build neural networks and get effective training. The graph structure in Pytorch is easy to understand, and more importantly, it is easy to debug for researchers. This article uses the train data set ,the accuracy of this network to reach 95.49%, and the image recognition has been able to have a higher accuracy.
❖ Train loss: 0.1732
❖ Train accuracy: 95.49%
❖ Test accuracy: 75.6% max