Food-Vision-101

• This uses various Deep Learning aspects and creates a model to predict 101 different classes of food.
• The dataset was taken from Official Food Vision 101 Research paper
• The official research paper has an accuracy of ~50% and our model has an accuracy of ~45%

Our Model

• I have made various models including curating my own model from scratch but that might not be sufficient as I have limited computation power
• So I have used ResNet models and finetuned them on our dataset

Model Summary before finetuning

Loss and Accuracy curves

Model Summary after finetuning

Loss and Accuracy before and after finetuning

Before (~33%) and after (~45%) finetuning Accuracy