Net Runner is doc.ai's open source, on device environment for running and evaluating computer vision machine learnings models. Doc.ai uses this tool internally to test our computer vision models in real world conditions, and we've decided to share it with the research community so that they may do the same.

Net Runner ships with five MobileNet image classification models that may be run on the CPU, GPU, or NNAPI accelerator. With the help of tensor/io, our cross-platform library for deploying TensorFlow Lite models to iOS and Android devices, researchers and developers can quickly test their models inside Net Runner to evaluate their performance in real world conditions.

Run inference in real-time on the camera feed or on still images, test your models on the CPU and GPU and under different multi-threading conditions, and, for developers, create custom UI to display the results of running the models you have added to the application.

Learn more about tensor/io at:
https://github.com/doc-ai/tensorio-android

With access to the complete source code for Net Runner at:
https://github.com/doc-ai/net-runner-android