We developed an easy and robust 3-step process to create a long term success for integrators and end-users.
A suitable 3D-camera is mounted to take extra images of the items in the actual picking environment.
Training of software
These images are enriched complementing our existing data set to train a neural network optimized for that application.
A locally installed PC analyses the images between each cycle of the robot. The optimized algorithm instantly creates the best possible 6DoF grasp poses including relevant information about the type of object.
Fizyr created algorithms for: