Edge Impulse: TinyML Language Classification Model
Madhura Patil, Prajwal Rawoorkar, Parth Muley, Sumitra N. Motade, Shweta Kukade, Anagha Deshpande, Arunkumar Nair · 2024
There is an immense demand for smart production procedures with a requirement of intelligence which is expected in today’s industry due to rapid development. One of those intelligent production techniques gave birth to TinyML, which has the ability to run machine learning models with low processing power and still satisfying the accuracy threshold for effective application. This research explores the implementation of TinyML for conveyor belt systems using Edge Impulse software. The significance of this research lies in the approach of handling a problem such as language identification on products real time and on edge devices. This pipeline was created with data acquisition, model design, training and testing done sequentially to suit our purpose, with getting the results of accurate identification of 4 languages Arabic, Chinese, Hindi and English along with processing time of 1ms and RAM consumption of 4KB. The study shares valuable insights about the practical implementation/ applications of Tiny ML in the smart manufacturing process by bridging the automated systems and human operations.