Android Hackathon Example

This project was made for the PyTorch Hackathon 2020, as a companion to a brief video showing how to stand up an Android project using the PyTorch Mobile runtime. You can view this video on the PyTorch YouTube channel: https://www.youtube.com/watch?v=O_2KBhkIvnc

Project Setup

To set this project up and run it in Android Studio, clone this git repository, and open the PTMobileWalkthruAndroid folder as a project. With the project open you should be able to build and run the project.

Model Preparation

The video describes a general process for getting your model into your project. If you'd like to try the pre-trained MobileNetV2 model used in the video, follow these steps in a Python environment with PyTorch 1.5 or higher and TorchVision 0.6 or higher installed:

1. python - open a Python REPL; the rest of the steps will be completed there
2. import torch and import torchvision
3. model = torchvision.models.mobilenet_v2(pretrained=True) will give you an instance of the model
4. Optional: Follow the model optimization procedure shown in the video
5. Required: scripted_model = torch.jit.script(model) will export your model to TorchScript
6. torch.jit.save(scripted_model, 'model.pt') saves your model to a file

For your convenience, copy the code below to test out the model creation:

import torch
import torchvision
model = torchvision.models.mobilenet_v2(pretrained=True)
scripted_model = torch.jit.script(model)
torch.jit.save(scripted_model, 'model.pt')

Project Notes

• This project is intended to be used with PyTorch 1.5 or higher, and Android Studio 3.5 or higher.
• If you want to upgrade the version of the PyTorch mobile runtime you're using, you can edit the build.gradle file for your app (not the file for the project).
• The first time you run the app and press the "Infer" button, it may take a while to load the model and run the inference. Subsequent inferences will be quicker.

For more information on PyTorch Mobile for Android, visit the PyTorch Mobile homepage.