Abstract: Based on video perception, using networks and information technology as media, and supported by artificial intelligence and big data technology, a full mine artificial intelligence (AI) supervision platform is constructed, and the overall architecture and core scenarios of the platform is introduced; based on the design of the entire platform, key technologies such as low sample dataset enhancement technology, model training technology, data reasoning and decision-making, and microservice backend development technology were discussed. The on-site application shows that the deep learning algorithm of the entire mine artificial intelligence (AI) supervision platform has an accuracy rate of over 90% in mining image classification, and an accuracy rate of over 80% in object detection.