Hi Folks, time for an update!

FLIR® Boson®

Besides for supporting the GenX320 Event Camera in the coming New Year we will also be bringing FLIR® Boson® support soon to the OpenMV Cam! With the FLIR® Boson® you have access to 640x512 Thermal Video at 60 FPS. While expensive, the FLIR® Boson® unlocks you to use the OpenMV Cam in high-end professional applications. In particular, you can buy the FLIR® Boson® in an awesome variety of configurations depending on your budget. FLIR® has a variety of different lenses available for the BOSON® to you can use it with any type of Field-of-View. You can also buy the BOSON® with or without a lens, reduce the resolution to 320x256, drop radiometric thermal calibration support, and pick between Professional and Industrial versions to save on the price.

Support for the FLIR® Boson® is made possible by a FLIR® Boson® adapter module which we will soon sell on our store. It works seamlessly like our FLIR® Lepton® Adapter Module by replacing the main camera sensor.

Additionally, with the RT1062 you can even create your own FLIR® Boson® PoE Camera (and FLIR® Lepton® PoE Camera too).

Cool. So, when will it be on sale? Answer - we hope to have the adapter in stock early in the New Year. We should be able to get it in stock before the GenX320 Event Camera Module we already announced.

Finally, if budget is no object (no kidding), you can even get yourself a Thermal Zoom system for the FLIR® Boson®.

Like ULAB? Want to make it go faster?

Our friend Zoltán Vörös is looking for help making ULAB (Numpy for MicroPython) run faster on ARM's new Cortex-M55 processors by leveraging the new Helium instruction set released by ARM. We've made ULAB a standard module on all OpenMV Cams and enabled 4D NumPy array support for all our boards. We're excited to support ULAB continuing to grow. Please read his call to action below:

Zoltán Vörös - Numerical acceleration under MicroPython

In the past, digital signal processing and machine learning applications on microcontroller platforms were bound by the numerical processing speed. Several years ago, ARM released the definition of the Helium extension to the Cortex-M CPU instruction set, designed to deliver significant numerical performance improvements of certain common operations.

On many variants of micropython, ulab, the numpy look-alike numerical library has been a popular choice for those, who want to retain the ease of python, but need to crunch numbers fast. The ulab project is planning on raising the bar and exploiting the advantages of ARM's Helium processors to accelerate many of its numerical methods. For this effort, we are looking for volunteers to accompany us on the journey. There are many ways of contributing.

1. You can dive into the Helium instruction set and work with us to develop high-performance implementations of key ulab functions, such as matrix operations, filtering, and more.
2. You can also help with benchmarking by testing the Helium-enabled functions and measuring performance gains, as well as by identifying areas for further optimization.
3. Ensuring that the new Helium-powered features are well-documented and thoroughly tested, thereby enabling users to easily take advantage of the improvements, would also be highly appreciated by the community.

If you are passionate about micropython in general, numerical methods, exploring the latest advancements in hardware acceleration, and you are proficient in C, we would love to hear from you. Even small contributions can make a huge difference, so do not shy away from reaching out to us at ulab!

OpenMV v4.4.7 Released

On a minor note, OpenMV IDE v4.4.7 has been released! The new IDE includes a number of bug fixes over v4.4.4. Otherwise, no major new features. However, GitHub deprecated Mac OSX Runner support for MacOS 12. So, OpenMV IDE v4.4.7 will be the last version built to run on MacOS 12. New Mac IDE versions will be built on MacOS 13. That said, we will still build the IDE for Intel to have universal Mac support.

Around the Web

Finally, Edge Impulse posted a cool tutorial showing off object counting using FOMO on the OpenMV Cam RT1062 - check it out!

Anyway, that's all folks!