Rapid development of embedded systems in recent years allows us to integrate powerful processing units in most electronic devices. Food preparation is one such field, where technology can ease everyday chores, and help us achieve better results. We've tried out various ways to use a camera in combination with artificial neural networks to control kitchen mixer when making whipped cream. First we made several recordings of cream during mixing. We've labelled each frame according to how well-mixed the cream is. We show some higher-importance neural networks for image analysis (AlexNet, MobileNet, NasNet) and test how well those neural networks perform after being trained on our dataset. Our results indicate that such neural networks are able to give an accurate prediction, even on photos captured under different conditions than the training data.
|
