Using Convolutional Neural Networks Libraries to detect and classify objects in industrial settings

Abstract

This paper summarizes an intern’s introductory observations about YOLO- a family of Convolutional Neural Network libraries currently used in Bangladeshi Artificial Intelligence Industry. A black-box approach is draped over the internal architecture of the model itself, and a greater focus is applied on observing the external factors such as input, output, metrics and the workplace environment that empowered the intern to study these factors. Specific test cases were designed to verify hypotheses about the model’s performance in specific situations. These verifications are used as further justification for the relevance of YOLO in the Computer Vision industry.