Good morning.
Place of integration:
Parts to detect by nano as :
Gripper:
Task to do: recognizing full layer of wooden parts on gripper, couse sometimes parts are sticked by glue together creating damages of gripper (robot takes partly two layers of parts and during squeezing two layers on convoyer makes damages with gripper, parts are taken by suction bars).
Using nano with 2d or 3d camera making check station. At the beginning collecting photos or point cloud mesh with operator decision “good/bad gripping”. this situation operator has to manually feed convoyer with parts taken from gripper and from layer of parts laying on wooden pallet.
Question 1: how about stability written in python application for (ml or 2d recognizing and via gpio connected nano with fanuc controller). Is it strongly recommend to use rtos ?
Question 2: which technology 2d or 3d is better to reach 100 accuracy. If 2d with ml which ml network use? I was only dig in in great explanated inference by Dusty. I have not much experience with training but it looks not so difficult with large dataset to label.
A propos labelling I have found provided by Hitachi online tool for labelling 2d or 3d
Can I label my dataset with it? Has not only rectangle labelling tool as in inference.
Question 3: which camera supported by Nvidia to use ? which 3d sensor to use ? I was thinking about L515 by realsense.
Question 5: where on market can I find board with relays and inputs to connect it with nano ?
Marek