Blog Presented Net Runner, the Device Environment for Computer Vision, at NeurIPS

Data scientists and machine learning engineers are the most in-demand professionals in the market to date. The downside of the growth of this community is the lack of ready-to-go infrastructure for easy and productive work. In order to address the needs of the community, especially researchers working with computer vision machine learning models, has released the Net Runner app for iOS.

The Net Runner project was presented at NeurIPS last week in Montreal at the “Machine Learning on the Phone and Other Consumer Devices” workshop.

Net Runner helps data scientists without mobile software experience prototype and evaluate models on mobile devices with TensorIO, that performs a layer of abstraction on top of an existing library, e.g. TensorFlow Lite; allowing it to focus on the input and output interface to the underlying model and to expose them in a declarative manner.

Now data scientists with a little knowledge of TensorFlow can add their models to the application and see the output in real time. In addition, models can be measured for latency and accuracy, run on hundreds or thousands of images in iPhoto albums, and be automatically tested.

TensorIO parses a description of the underlying model in a JSON file format and uses platform-specific APIs to perform the required image operations. Since the declarative specification is plain text and platform independent, it can use language that most machine learning engineers are familiar with — types, shapes, and transformations.

In conjunction with a JSON file, a TensorIO bundle requires a model file and an optional assets folder. We are focused on TF models, but TensorIO is built to support other machine learning frameworks in the future.

At, we have gathered a team of the most experienced AI researchers and engineers under the leadership of our Chief Science Officer Jeremy Howards with the aim of building the most advanced tools for the community. Creating Net Runner and TensorIO along with a decentralized infrastructure for mobile devices is the first step towards that goal.

Cutting-Edge Machine Learning

Today, we are witnessing that computing is centralized and executed in the cloud powered by tech giants like Amazon, Microsoft, Google, and IBM. With that approach, privacy and security can be at risk. With our concern for that issue and our conviction that machine learning will be moved onto mobile for the sake of speed and security, we created infrastructure for devices to advance data scientists’ work with models and contribute to the community and spirit of privacy.

Consider the clear on-device advantages:

Speed. It can take up to 20 seconds to run a model in the cloud, but if you run the same model on a device, it takes a few milliseconds to significantly reduce latency. The model can also be tested offline to bring even more convenience to your work.

Privacy. By working on a device, you won’t share your data with third parties.

Security. is driven by the mission to bring decentralized machine intelligence into the market to ensure much higher security in your work. Moreover, the concept of building decentralized computing along with a learning ecosystem has been on the minds of the team since the founding of the company and is reflected in our brochure and white paper.

Check out the Net Runner GitHub repository along with TensorIO, or download the app from the App Store, test the code and let us know what you think. Just keep in mind that Net Runner currently supports TensorFlow Lite models and works on iOS 9.3 or higher.

We are looking forward to hearing from you and learning how we can improve our tools to make your work faster and more productive!