VGG Fine-tuning for Cooking State Recognition

TL;DR

This paper presents a fine-tuning approach for the VGG CNN architecture to accurately recognize eleven cooking states of food ingredients, aiding domestic robots in cooking tasks.

Contribution

It introduces a fine-tuning method for VGG CNNs specifically tailored for cooking state recognition, achieving around 76.6% accuracy on validation and test sets.

Findings

Validation accuracy of 76.7%

Test accuracy of 76.6%

Optimized model parameters for best performance

Abstract

An important task that domestic robots need to achieve is the recognition of states of food ingredients so they can continue their cooking actions. This project focuses on a fine-tuning algorithm for the VGG (Visual Geometry Group) architecture of deep convolutional neural networks (CNN) for object recognition. The algorithm aims to identify eleven different ingredient cooking states for an image dataset. The original VGG model was adjusted and trained to properly classify the food states. The model was initialized with Imagenet weights. Different experiments were carried out in order to find the model parameters that provided the best performance. The accuracy achieved for the validation set was 76.7% and for the test set 76.6% after changing several parameters of the VGG model.