The aim of this project is to recommend recipes based on a query image of a dish that is provided as an input. This project involves following modules:
- Data Preprocessing : All the images in the food image database and the query image is reshaped to the size 2241693.
- Feature Extraction : Features are extracted using feature extraction techniques like KAZE, SIFT, SURF and HOG.
- Ensembling Features : Here, all Features obtained from above feature extraction techniques are combined/stacked.
- Feature matching : Ffeature matching is performed between the query image and the images of the predicted class using cosine similarity.
- Class Prediction : Various classifiers are used to predict which class the query image belongs to.
- Image and Recipe Retrieval : First 5 images and their recipes are retrieved based on the least distances or maximum similarities.
- UI : These dishes and their recipes are displayed in a website using HTML and server is hosted using Flask.
