Mitsui Chemicals Deploys AI and Automation Systems with TensorFlow and MATLAB

“MATLAB solved our problems on the field implementation and saved development time. That led to highly accurate development.”

Key Outcomes

  • Reduced visual inspection time by 80% 
  • Effectively used models trained in other frameworks
  • Deployed application with a user interface that anyone can use

Mitsui Chemicals develops factory automation solutions by applying AI, mainly machine learning.

At the beginning of development, Mitsui Chemicals used Python+Keras (TensorFlow) for automated visual inspection of sheet-shaped products on the production line. However, ease of use and maintenance were issues in implementing the trained models in the field.

Mitsui Chemicals engineers chose MATLAB® to create applications with easy-to-understand user interfaces. The model was imported using the Deep Learning Toolbox™ Importer for TensorFlow-Keras Models. The created application was distributed and executed using MATLAB Compiler™ and the MATLAB Runtime.

By using MATLAB, Mitsui Chemicals engineers solved the challenges of field implementation, reduced development time, and achieved highly accurate development.