Coffee Bean Project

About The Project

Our key project focuses on the development of specialty coffee bean sorting, using artificial intelligence particularly deep learning and pattern recognition to identify, recognize and distinguish objects as tiny as coffee beans from different angles, and be able to classify and sort them based on quality, grading or other requirements tailor made for our prospective customers. 

Quality Inspection

Classification of coffee bean

Our network can distinguish different kinds of defects 

Here is the different classes of coffee beans that our network can distinguish out.

Semantic Segmentation

How It Works?

  • Our network can distinguish out insect damage, coffee beans and background pixels by pixels.

  • Insect damages are in grey, coffee beans are in white and background is in black.

  • The following images show the result of our network. The insect damage can be distinguished by our network.


Good Bean

Insect damage

Class Activation Mapping

From the image, we can know how the neural network do inference. Color closer to red means the network is mainly using the features around that part to do inference.