Deep learning industrial vision for glass bottles

Deep capture, machine vision software in deep learning allows the detection of bottles position (X, Y, Teta) in different types of containers dynamically on a conveyor.

The system is integrated in a robotic depacker which allows the taking of the bottles for a filling line loading.

The difficulty of detection lies in the diversity of the bottles references to be treated (around fifty) and of the containers (cardboard boxes and thermoformed trays).

Deep capture makes debugging and creating formats easy. A single lights and cameras configuration allows all references detection. An algorithm makes it possible to order and return the gripping positions to the robot.