Inspiration
We like collecting cards and want to determine which cards we want to grade as grading is expensive. Currently, on the market, there is a hole for an app that tells you how well a card will grade. In our project, we sought to fill this niche in tandem with our desire to learn skills from industry.
What it does
The user can take a picture on the front page and back page of the card. The app will send these pictures to the backend and input them into the CNN-supervised ML model to predict the grade level and price.
How we built it
We built an Android app using the react-native framework to serve as our frontend UI/UX. The data is scrapped using selenium, the image will then be preprocessed using opencv, and finally fed through a model built using tensorflow. The information will then be packaged via an API built using flask and sent back to the frontend.
Challenges we ran into
Collecting and scrapping large amounts of data efficiently was really difficult. Preprocessing data in order to allow the model to make accurate predictions was a challenge we had to navigate. Lastly, the model itself was difficult to train and attain high accuracy.
Accomplishments that we're proud of
We are proud that we managed to complete the project and finish the integration of frontend and backend.
What we learned
Data is hard to come by, ML models can solve a lot of challenges given the right tools, having a good understanding of the mathematics behind the code is important.
What's next for PokeCardtel
Expand the model and collect more data. Allow more than one user and have better UI/UX design.
Log in or sign up for Devpost to join the conversation.