The shazam for food.
This app is loosely based on Silicon Valley's SeeFood
-
I trained a neural network using Keras and Tensorflow to export it as a tensorflow lite model.
-
Conducted hyperparameter tuning using Weights and Biases.
-
Implemented the back end in Kotlin.
-
The app supports classifier for hot dogs, but one could increase the scope of the classifier with a hierarchical softmax loss (only if the number of categories exceeds thousands.)
-
The result of the classifier is uploaded to the screen along with the image that is classified.
Here is a video implementation:
