TensorFlow, OpenCV & custom Dobot APIs for sorting raw & ripe tomatoes using Deep Learning Robot

Robotic binning of raw & ripe tomatoes using Deep Learning, TensorFlow, OpenCV & custom Dobot APIs. The processing was done on Intel Core i5 8th Gen processor and NVIDIA GEFORCE MX150 graphics processor. Custom APIs were written to program the Dobot Magician robot arm. TensorFlow, OpenCV and Python were integrated to accomplish the task at hand.
Jig for webcam mounting was designed, 3D printed, assembled and Robotic programming was done at Assemtica Robotics.
For more information, please visit our website - assemtica.com/
Design hardware: Apple iPad 2018 and Apple Pencil
Design software: Shapr3D
3D Printer: Flashforge Inventor II
Webcam: Logitech C310
Robot: Dobot Magician
Conveyor: Dobot

[@renzojacobo8509]

Could you share the book, paper or data that you used to build this projects thanks from Perú

[@arafat877]

Could you share with me the technical design, and software of this robot please ???

[@sonalishashikantrokadecomp9896]

can you share me details of connectivity and programming dobot magician using jupyter notebook. I am not able to connect it.

[@silcaloutube]

Can this be adopted to a real industrial application with industrial grade robots?. Using a Logitech webcam concerns. Wat would be the influence of ambient lighting in this demo.

[@selineni]

We also use basler ace2 cameras with higher mega pixels for industrial cases. Since wee are using mask rcnn there is little to none effect of ambient light

[@pavan3046]

Can you share the code
Or explain us how to build it.
Would love to build something like that

[@selineni]

Akash. Do you have a robot? It is done with custom dobot api and tensorflow. Some of it is available on my github pagegithub.com/seli07 if you use the code please give me a star in github. Thanks

[@vittalrathod6536]

How to connect Rahul sir please let me know

[@pradipubale1985]

Very nice

[@Kanjox]

awesome

[@Yourself98]

hi guy, can you help how to send data from laptop to robot? thanks

[@selineni]

The overall process is easy. Only complication is mapping localization information from tensorflow or pytorch to robot. We can do that either with a linear mapping of camera pixels to surface or you can use dot product for projection of camera axis to surface axis. You can use an arico marker for hand eye calibration

[@micah6635]

pog

[@micah6635]

michael reeeeeeeeeeeeeeeeeeeeeeeeeeee

[@williamhuang5329]

Hanzhen harmonic gear , robot gear reducer, over 30 years experience

[@jserra1176]

weak