Android Application using Lazy CNN on the Example of Digit Detection

The following paper focuses on the number of questions; describes the approach of deep-learning in the goals of mobile calculations. The paper describes the approach of convolutional neural networks using modified architecture so-called “lazy convolution”. The results showed that this approach helps to minimize computation complexity. This work contains an introduction part, three chapters, conclusion part, references, and appendixes. The first chapter brings an overview of a problematic area, it covers related works and research background. The second chapter provides proposed method idea, algorithm details, limits, and specifics. The third chapter presents statistics and accuracy results from algorithms implementation, gives android architecture details, transportations graph for android. The work consists of 39 pages, includes 14 figures and 7 tables, 8 listings and 3 appendixes.