Computer Vision for Quality Control: Buckwheat, Defects, and a Free Platform for Labeling and Model Training

The article explores the practical application of computer vision for automating quality control on production lines. The author demonstrates the process using a buckwheat sorting system designed to detect defects. A key feature of the solution is the use of a free platform for data labeling and model training, which significantly lowers the barrier to entry for small businesses. Special attention is given to ease of deployment: the final product is a compact HTML file containing the pre-trained model. This allows the system to operate autonomously without an internet connection, which is critical for industrial environments. The author shares their experience in developing and optimizing the model, highlighting the accessibility of tools for building effective machine vision systems without significant infrastructure costs.