Are you sure you can drive? Know it with this simple but inaffective tool 😅.
I developed this app with a focus on mastering the Core ML and Create ML APIs.
The app presents challenging math questions, requiring users to provide their answers to assess their sobriety. To achieve this, I harnessed Apple's MNIST model, which recognizes handwritten digits. Moreover, I integrated PencilKit to offer users a canvas for drawing their responses.
![](https://private-user-images.githubusercontent.com/28783605/268442768-c3a28f00-3e69-4672-b007-4b53e96c3fa7.gif?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.lQFyEfGnvJBF38w7RRioA5TVsOhY4hy2gNsBlQeC7tA)
During my learning process, debugging the ML model posed a significant challenge. Initially, the model's accuracy was subpar. To improve it, I addressed two pivotal issues: a) modifying the input format to white on black and b) increasing the marker width on the canvas.
Used util classes from: CoreMLHelpers