Design And Implementation Of Digit Recognition System Based On Deep Learning

Image recognition technology plays an important role in many fields, while digit recognition is an important branch of image recognition. Many applications of businesses require to recognize the digit information from images accurately. Improving the speed of digit recognition can greatly enhance the efficiency of small and micro enterprises.Traditional digit recognition approaches extract feature from images manually that lead to deficiency and low-robustness of the algorithms. Compared with the traditional methods, feature extraction with deep learning technology, which has become one of the most popular research, no longer requiring human intervention, as to ensure a faster speed. Based on deep learning, the paper design and develop a system to solve the problem of digital recognition and help to prompt the efficiency of the recognition algorithm. First, this paper describes the present studies on deep learning and image recognition. Second, it introduces deep learning framework for digital recognition and related training tools, models, transporting technologies and image preprocessing methods. Then, the overall structure and detailed design of the system is introduced. Finally, data samples are collected to test the entire system to analyze the accuracy and efficiency of the proposed approach.