The thesis explores the use of computer vision in the industry, with a special focus on convolutional neural networks (CNN). In the introduction, it discusses the history and current trends in these areas. The work examines in detail the challenges and processes of learning neural networks for industrial computer vision, including the methodology of learning, model evaluation, network architecture, their results, and practical implementation. Furthermore, the development and implementation of a computer vision model in an industrial environment are presented. This includes the collection and processing of image data, modeling, labeling, learning, optimization, API development, integration points, documentation, and training. The thesis also analyzes and compares different approaches and tools for learning computer vision models, such as Microsoft Custom Vision, Roboflow, and YOLO, and focuses on the importance of data capture, labeling, and processing for the success of computer vision.
|
