Dish Detective

Recipe Page

Inspiration

Growing up, I was always in the kitchen. Whether it was bothering my mom to make my favorite meal or learning to cook it, food has always been a part of my life, and I love to cook. Recently, I've noticed I'm losing my passion for cooking. A lot of this comes from having too many or too few recipes. I don't always have the ingredients to make what I see online, and I end up spending more time scrolling than actually cooking. Some days, I just want to come home and know what to make, and I realized it would be great to have the recipe based on my ingredients rather than the other way around!

What it does

This app allows the user to upload a picture of fruits and vegetables they want to cook with, and then it automatically generates a recipe. You would then be able to save, screenshot, or share it with friends!

How I built it

I first thought to use TensorFlow to read the objects but then moved to Google Cloud Vision. They have an API that works with Python to process an image, give a confidence score, and provide a list of what the program sees in the picture.

After the fruits and vegetables are extracted, we all ChatGPT 3.5 via API to generate a recipe.

This is then shared on the website (using FLASK to host the Python side and ReactJS to host the website). The user can upload their own picture, and the image is processed with the program again.

Challenges we ran into

I had quite a few issues with this project that I had to overcome. Firstly, I had this really big idea, but quickly had to narrow it down to have a product done by the time of the submission. I really had to think of how to scale it down, and it was very challenging. I then fell into a rabbit hole with computer vision. I got extremely interested and lost in the intricacies of how it worked, reading article after article before realizing I had spent too much time researching. My last issue was not knowing how to do well, ANYTHING for the project. I watched many tutorials and asked for help from the others around me and figured out the python aspect, the backend. I then spent a while deciding on a theme and finishing a mockup for a website.

Accomplishments that we're proud of

I really am proud of how much I learned in the process of developing this idea. I have a MUCH better understanding of computer vision and how the computer processes images. I also feel like writing and understanding code that uses API calls, tokens, and keys has helped unlock a lot more potential for things I can write in the future.

What we learned

On a technical level, I learned how computer vision works, how to design webpages, and how to use API calls effectively. The most important lesson I learned was keeping scale in mind when developing. I had to learn to resist adding many features at once, and focus on having a polished smaller product.

What's next for Dish Detective

In the short term, I want to be able to connect the front end and back end, having a full website done. As for the long term, I plan on adding mobile app support, I feel like many people would genuinely use this product, and having it in the palm of their hands would increase use massively. As for the LONG long term, with the development of AI/ML software, we will see, much smarter recognition software, and eventually the program will include things such as seasoning, canned goods, and cartons.