Clasificador de imágenes y entrenamiento en la nube

Abstract

Nowadays, the most used techniques for performing object classification in the field of Computer Vision are based in Deep Learning, mainly Convolutional Neural Networks. The process of training the network often takes a long time to achieve acceptable results, the field dedicated to training the models have seen a fast evolution in this last decade on pace with the evolution in the algorithms themselves. The use of GPUs and Cloud Computing have been the more notable of these advances. In this way, Cloud Services Providers have seen the opportunity in normal users or small companies who lack the computing power needed to train large, Deep or Machine Learning models. They offer them the possibility of using their computing power to achieve this task in an affordable training time, usually by renting the servers in one way or another. In my project I will go through the process of training an object classifier in the Could Services offered by Google, explaining both the technology backing the Neural Network and giving an insight into the world of Cloud Computing applied to the training of Neural Networks.