swaad (hindi for 'taste') follows our belief that leftover food just needs a little redsign and it can save a lot of food going to waste. swaad is a simple-to-use, cost-effective application that requires you to take pictures of left-over items like fruits and groceries, and it will use Computer Vision algorithms to identify them and suggest recipes to make from them!
The UN's Sustainable Development Goals Action Campaign is a cooperative effort to develop a sustainable thinking and tangible approaches towards sustainability. In 2018, The Chef's Manifesto was one of the winners of the UN SDG Action Campaign Award, to create a more holistic approach toward's sustainability, from the perspective of food. Since food production cannot be avoided, it is important to reduce it's wastage. But the question is, can we do something similar at home?
In the U.S., up to 40 percent of all food goes uneaten each year, at an annual cost of $218 billion!
What it does
- Android app that is easy-to-use and understand
- No special electrical/hardware requirements, just click pictures of food!
- Segmentation and Recognition of more than 250 fruits and vegetables
- Select multiple leftover ingredients at once
- Smart recipe recognition - Recipes ordered on the basis of how many new ingredients do you wish to purchase!
How we built it
- Android application with the help of - Android Studio
- Image segmentation and recognition with the help of - AWS Rekognition
- Simple RESTful API backend server - Flask
- Food recipe recommendation with the help of - Food2Fork
Challenges we ran into
- Deciding a way to present the solution in a better manner. Sustainability is a very difficult theme to think on (compared to Education and Wellness), and we're glad we took up this challenge!
- Finding and training vision models that work decently enough on food data. Most of the libraries we found turned out to be a dead-end after 2-3 hours of investment. Ultimately, we made use of AWS Rekognition.
- A common struggle with every application ever - networking! It requires a lot of domain knowledge and sending image files from android to the flask backend can be tricky!
Accomplishments that we're proud of
Building a fully-functional, useful app in less than 36 hours! Some of the tech stack we had never used before. Some of the errors and technical nitty-gritties set us back from our regular course, but we came through stronger! We learned a lot of computer vision algorithms, the importance of image compression and how android studio sometimes doesn't work at all :/ Also, the fact that we can now recommend recipes with ingredients mixed and matched in the weirdest combination really entertains us
What we learned
- Using technology for social good should be thought over. Sometimes important challenges are very solvable!
- Network debugging/implementation can be tricky
- Sleep is the most important thing ever, please do not ignore it
What's next for swaad
- Add a view/page to show the current list of scanned ingredients .
- Expansion to more food items like cooked food and spoilt food .
- Live usage in households and benchmarking food wastage before and after regular use of the app!