Machine learning for IT

Expert systems and knowledge bases

About this app

The ML_IT app can be used to train users in machine learning techniques on well-known datasets such as Fashion MNIST, MNIST, and Cifar-10. A case study is presented for creating a neural network for Fashion-MNIST images. The user can compare the neural network's image prediction results with those of the Tflite model. Note that due to the small data sample, the neural network's results are slightly worse (80%) than those of the Tflite model (95%). The user can predict a digit from 0 to 9 by drawing it on the smartphone screen with their finger. It is also possible to edit the digit's writing with a finger to obtain a better prediction. The Tflite model is used to predict images based on the Cifar-10 dataset.

What’s new

This version of the ML IT app adds the ability to load the phone's front and back cameras for face localization. Face localization in streams is also available. The Tflite model is used for face recognition.