qiubit / traffic-sign-android

Traffic sign classifier trained using fastai and deployed to Android using PyTorch Mobile

traffic-sign-android

Traffic sign classifier trained using fastai and deployed to run in real time (60-70 FPS) on Android using PyTorch Mobile.

The Android app is based on pytorch/android-demo-app, an official demo app for PyTorch Mobile project.

Training the model

To train the model, launch Traffic_Sign_Classifier.ipynb notebook using Google Colab, preferably using GPU runtime.

This repo can be directly imported using Google Colab GitHub import feature.

Building and launching the app

To build and launch the Android app, open Android/build.gradle in Android Studio.