Going to the store and getting getting a bad apple, a grainy watermelon, or poor quality produce can very easily ruin your day. Furthermore, this increases the food waste that groceries end up creating. This unfortunate event occurs your options dwindle your options result to just wasting your hard-earned money you spent on the produce. Simultaneously it wastes our Earth's resources and people's time.
What it does
Our app is a early level implementation of a quality of food detector to help distinguish between good and poor quality food to help consumers make the right decision on the produce they will buy. Furthermore, it can help grocery stores quickly filter and decide when to sell their produce such that it is optimized where no food is unnecessarily wasted. Users can simply put the item under the camera of their phone and then the machine learning model will output whether or not the item is fresh, approximately how long it will last, and how confidant the model is in its prediction.
How we built it
We used machine learning model through Google's cloud API we were able to train a model that could help detect fresh vs spoiled food.
Our tech stack
Challenges we ran into
Finding enough pictures to train a good model was difficult. With more time and access to other sample data it would be easier to get more pictures and as such result in a better model.
Accomplishments that we're proud of
The model at a basic level works well and could be used to distinguish things, especially apples.
What we learned
We learned more about machine learning models and how the data we feed it can affect end result we are looking for. We learned about how machine learning models work and the difference between supervised and unsupervised learning, and put these to practice. We also tried implementing this rudimentarily with Tensorflow but due to a lack of data we found it very difficult.
What's next for fresko
Train, train, train. We would like to train the model much more on various things. These include on more types of fruit and products such as cheese. We would also like to help it detect more niche differences between ripeness levels to provide a stronger use need so that people can be more precise when their food will spoil. A mobile app would also allow us to give users much more accessibility to use our platform.