Enhancing Food Results Accuracy

Our current implementation leverages GPT-4 for extracting both ingredient names and their nutritive qualities from images of food items. However, this approach sometimes leads to inaccuracies in nutritional data due to the generalized nature of GPT-4's outputs. To improve the precision of caloric value determination, we propose to refine our process by exclusively extracting ingredient names using GPT-4 and subsequently querying a dedicated nutritional database or API for precise caloric values.

Objective

• Primary: To solely extract ingredient names from food item images without their nutritive qualities using GPT-4.
• Secondary: Identify and integrate a reliable nutritional database or API that provides accurate caloric values for a wide range of ingredients.

Tasks

1. Research and Evaluation:

   • Investigate potential databases or APIs specializing in nutritional information.
   • Evaluate the candidates based on data comprehensiveness, accuracy, API response times, and cost-effectiveness.

2. Implementation:

   • Adapt the existing GPT-4 integration to focus on the extraction of ingredient names only.
   • Develop a module to query the selected nutritional database/API with the extracted ingredient names to obtain precise caloric values.

Start by locating the existing GPT-4 image-processing integration in the repository and documenting how ingredient and nutritional data are currently extracted. Evaluate candidate nutritional APIs against coverage, accuracy, response time, and cost, then define completion as ingredient-only extraction plus accurate caloric lookups through the selected API.