Skip to main content

How to run ML Applications on Particle Hardware

With the release of TensorFlow Lite at Google I/O 2019, the accessible machine learning library is no longer limited to applications with access to GPUs. You can now run machine learning algorithms on microcontrollers much more easily, improving on-board inference and computation.

[Brandon Satrom] published a demo on how to run TFLite on Particle devices (tested on Photon, Argon, Boron,  and Xenon) making it possible to make predictions on live data with pre-trained models. While some of the easier computation that occurs on MCUs requires manipulating data with existing equations (mapping analog inputs to a percentage range, for instance), many applications require understanding large, complex sets of sensor data gathered in real time. It’s often more difficult to get accurate results from a simple equation.

The current method is to train ML models on specialty hardware, deploy the models on cloud infrastructure, and backhaul sensor data to the cloud for inference. By running the inference and decision-making on-board, MCUs can simply take action without backhauling any data.

He starts off by constructing a simple TGLite model for MCU execution, using mean squared error for loss and stochastic gradient descent for the optimization. After training the model on sample data, you can save the model and convert it to a C array for the MCU. On the MCU, you can load the model, TFLite libraries, and operations resolver, as well as instantiate an interpreter and tensors. From there you invoke the model on the MCU and see your results!

[Thanks dcschelt for the tip!]



from Hackaday https://ift.tt/2QVUtfY
Labels:

Comments

Post a Comment

Popular posts from this blog

Bill Gates steps down from Microsoft’s board to focus on philanthropy

In an announcement on Friday, Microsoft revealed that company co-founder Bill Gates has decided to step down from his role on its Board of Directors in order to focus on his philanthropic efforts at the Bill & Melinda Gates Foundation. This is Gate’s biggest change to his role at Microsoft since stepping down as company chairman in February 2014. According … Continue reading from SlashGear https://ift.tt/2We90Gu
Post a Comment
Read more

How To Play Doom – And More – On An NES

Doom was a breakthrough game for its time, and became so popular that now it’s essentially the “Banana For Scale” of hardware hacking. Doom has been ported to countless devices, most of which have enough processing ability to run the game natively. Recently, this lineup of Doom-compatible devices expanded to include the NES even though the system definitely doesn’t have enough capability to run it without special help. And if you want your own Doom NES cartridge, this video will show you how to build it . We featured the original build from [TheRasteri] a while back which goes into details about how it’s possible to run such a resource-intensive game on a comparatively weak system. You just have to enter the cheat code “RASPI”. After all the heavy lifting is done, it’s time to put it into a realistic-looking cartridge. To get everything to fit in the donor cartridge, first the ICs in the cartridge were removed (except the lockout IC) and replaced with custom ROM chips. Some modifica...
Post a Comment
Read more

A Mini USB Keyboard That Isn’t A Keyboard

A useful add-on for any computer is a plug-in macro keyboard, a little peripheral that adds those extra useful buttons to automate tasks. [ Sayantan Pal] has made one, a handy board with nine programmable keys and a USB connector, but the surprise is that at its heart lies only the ubiquitous ATmega328 that you might find in an Arduino Uno. This isn’t a USB HID keyboard, instead it uses a USB-to-serial chip and appears to the host computer as a serial device. The keys themselves are simple momentary action switches, perhaps a deluxe version could use key switches from the likes of Cherry or similar. The clever part of this build comes on the host computer, which runs some Python code using the PyAutoGui library. This allows control of the keyboard and mouse, and provides an “in” for the script to link serial and input devices. Full configurability is assured through the Python code, and while that might preclude a non-technical user from gaining its full benefit it’s fair to say that ...
Post a Comment
Read more