Convolutional Neural Network (CNN) för klassificering av ritningar i dokument

Sammanfattning: This bachelor’s thesis was implemented with the purpose of investigating the development and implementation of a classification model for identifying drawings in PDF files. The thesis has been conducted at Tendium AB, a company that aims to simplify public procurement using artificial intelligence. Two different models were created with varying dataset sizes to evaluate performance. The models were constructed as CNNs with multiple layers. The study aims to evaluate the feasibility of the implementation, how it can reduce manual work and contribute to increased value proposition, as well as the benefits it can bring to Tendium’s positioning in the public procurement market. The results demonstrate that the model has excellent performance metrics and that implementing it in Tendium’s business can contribute to an enhanced differentiation strategy based on Porter’s generic strategies.