OnnoMitro
Personalized Bangladeshi meal recommendations from a new 305-recipe dataset
OnnoMitro (“food friend” in Bangla) is a web application that recommends personalized daily meals from traditional Bangladeshi cuisine. Most meal recommenders are built on Western datasets, which makes their suggestions hard to follow in Bangladesh.
What we built
- A new Bangladeshi meal dataset: 305 recipes collected by web scraping from three Bangladeshi recipe websites, tagged by meal type and vegetarian status, with per-serving calories, carbohydrate, protein, fat and fiber calculated using the Edamam Nutrition Analysis API
- A recommender system that estimates each user’s energy needs from their metabolic rate, activity level, BMI and goal, splits them into macronutrient targets, and ranks recipes by a weighted nutrient-distance score
- A Flask web application with sign-up, a personal dashboard, daily meal plans, alternative suggestions and calorie tracking
The OnnoMitro app: dashboard, daily meal plan and nutrition overview.
Results. On 50 test users, daily recommendations differed from the target calorie intake by a median of 22 kcal (mean absolute error 85.84 kcal). The largest errors came from the uncommon case of users with normal or high BMI who want to gain weight.
Code on GitHub · Senior Design Project (CSE 499), North South University, 2024 · with Tasin Tayeba Khan and Moshiur Rahman, supervised by Dr. Sifat Momen