Our group hopes to implement a deep learning model to classify food images with high accuracy by using CNNs to generate this network. We are using the Kaggle dataset of images and captions called “Food-101.” Inputs would be food images and outputs would be the label of the food image. Our challenges thus far included finding the proper dataset for our project and increasing accuracy. The initial challenge with this project was finding the proper dataset as our initial choice held around 1000 train and test images which was not enough for proper training or testing, especially considering our project target goal. We then changed our dataset to “Food-101” which includes 5GB of food images with labels that we could use. The next issue was increasing the accuracy of our model. We are basing our current work off the paper “Food Classification from Images Using Convolutional Neural Networks” which describes in detail model details, but we are struggling to increase accuracy. With the left on this project before DL Day, we will have more time to increase the accuracy of the dataset. We are currently on track to continue increasing accuracy now that our paper and dataset are finalized. We are looking to add more CV or DL elements to our model, with the focus on reaching our base goal by the end of this week and our target goal by the presentation day.
Log in or sign up for Devpost to join the conversation.