Ingredient Detection, Title and Recipe Retrieval from Food Images

With advancements in the machine learning community and an increase in social media usage in our daily lives, we see a steady growth in the number of food pictures shared online by users on a daily basis. Unfortunately, the picture does not adequately convey the intricate recipe that was used to prepare it. Thus, the image-recipe association problem is addressed in this study. Large datasets like Recipe1M+ and Food101 were accessible for our use, however, they do not rank highly for Indian dishes. As a result, a unique dataset that is entirely devoted to Indian cuisine has been developed and utilized. The aim is to develop a website that can take an image of an Indian dish as input, identify its ingredients, and retrieve the dish’s title. Further, this title is used to retrieve the recipe from the web. The task can be extremely challenging for a computer since it must understand the data well enough to produce accurate and precise ingredients. Both the dataset and the code will be made publicly available.