Copyright © 2020 AI Venture Labs. All Rights Reserved
STARTUPS DETAILS
eyecan.ai
Our solution is complex, but basically we can convert any industrial robot (also collaborative robots), equipped with a camera, into a robot capable of training itself (or even other robots) to recognize any type of object (2D or 3D pose, pixelwise segmentation, depth prediction) without the need to write new software!
Description
Artificial Intelligence applied to images today is able to work miracles. It can detect skin cancer, recognize the face and movements of thousands of people at once, generate completely fake videos that are indistinguishable from real ones, and even let cars drive themselves! In the industrial field it can help to recognize any type of object and also to identify its defects. But all this is not free! AI is a cost because you need hundreds of thousands of Training Images to make it work (and these are now generated manually by humans). Training an AI-based Industrial Vision system would therefore cost me too much because the time needed to produce the data would be excessive compared to classical methods. EYCAN offers a secret sauce: we developed a Robotic (patented) solution capable of generating hundreds of thousands of Training Images without human intervention in few minutes! This approach is 1000 times faster than the manual approach and produces perfect data (manual data contains errors) in a confidential manner (using labeling service for industrial data often violates non-disclosure agreements!).