Dataset expansion and accelerated computation for image classification: A practical approach

Abstract

The training dataset of many machine learning algorithms for various purposes mainly consists of images. The major hindrance and setback during the training of these datasets arises in the form of non-availability of the following three features - quantity of data, availability of GPUs (Graphic Processing Units) and high-rate computation catalysts. Many researchers have trouble independently training datasets and specifying features which can be in great quantity for images. In this paper, we present an approach for leveraging the power of “transfer learning” and easily accessible examples in the form of raw content from the internet not only to use already-prepared datasets made specifically for neural network training but also to bring into usage more training examples using the internet, sampling the average accuracy output rate of the images, along with reducing model training and execution time by parallel operations on different nodes.