Kitchen Utensils Recognition using Fine Tuning and Transfer Learning

To support blind persons at home especially in the kitchen, this work proposes the recognition of kitchen utensils using video sunglasses. The recognition system is based on transfer learning/fine tuning an existing deep learning algorithms, VGG16. Initially, our system can recognize 6 kitchen items using 1354 images in 6 classes. The training/validation and evaluation sets are set at 80% and 20% respectively. Most of the training data was downloaded from the Internet. In this challenging and noisy data, we achieved and accuracy of 95% using the fine tuning learning.