Machine Vision with MicroPython and the OpenMV H7 Camera Board

I’ve talked about OpenMV’s H7 camera board before when I’ve discussed the rapid growth in the use of machine learning on low-powered edge computing devices. Back then, towards the middle of the year, the hardware itself wasn’t yet available. But that has changed, with the OpenMV H7 now raising on Kickstarter.

The OpenMV H7 is an open source machine vision camera designed for low-power, real-time applications running MicroPython.

The OpenMV H7 board, front (left) and back (right). (📷: OpenMV)

The board is built around a 32-bit STMicroelectronics STM32H743VI ARM Cortex-M7 processor, with 1MB of RAM and 2MB of Flash, running at 400MHz.

The OpenMV H7 pin outs. (📷: OpenMV)

The new H7 board is the successor to OpenMV’s previous camera board, the M7. While a solid board, the older M7 camera board couldn’t run the machine learning algorithms that have recently become so interesting on the edge.

The new board comes with an OmniVision OV7725 image sensor is capable of taking 640×480 8-bit grayscale images or 640×480 16-bit RGB565 images at 60 FPS when the resolution is above 320×240 and an impressive 120 FPS when it is below. However, interestingly the image sensor is removable, and comes with a 2.8mm lens on a standard M12 lens mount. This means that both the lens and vision sensor itself can be swapped out and replaced.

OpenMV is offering two alternative images sensors, the Global Shutter Camera Module intended for “professional machine vision applications” and a FLIR Lepton Adapter Module for “thermal machine vision applications.”

The Global Shutter Camera Module. (📷: OpenMV)

The Global Shutter Camera Module is based on the ON Semiconductor MT9V034, a grayscale sensor capable of running at 80 FPS in QVGA mode, 200 FPS in QQVGA mode, and up to 400 FPS in QQQVGA mode.

The FLIR Lepton Adaptor Module is exactly what it says on the tin, just an adaptor board, as the actual FLIR Lepton sensor is sold separately due to export control restrictions.

The FLIR Lepton Adapter Module, with FLIR Lepton sensor mounted. (📷: OpenMV)

However, OpenMV is also working to support both the Panasonic AMG8833 GridEye Thermal Sensor and Melexis MLX90640 Thermal Sensors with the same adaptor module, which should give users access to thermal imaging support at both a much lower price point, and without the restrictions of US export controls.

The OpenMV H7 is currently on Kickstarter, with price points ranging from $49 for a camera board with the default pre-focused OV7725 camera module mounted, through to $99 for a board with a mounted OV7725 module and an additional the MT9V034 camera module included, all the way up to $329 for a board with a mounted OV7725 camera module along with a FLIR Lepton adapter module, and a FLIR Lepton 3.5 camera module included. Default shipping is via Hong Kong Post, although you can pay extra for upgraded shipping via FedEx, DHL, or UPS.

While OpenMV intends to make the board schematic, and EAGLE board files public, they won’t be available until after the Kickstarter campaign concludes.

Machine Vision with MicroPython and the OpenMV H7 Camera Board was originally published in Hackster Blog on Medium, where people are continuing the conversation by highlighting and responding to this story.

Original article: Machine Vision with MicroPython and the OpenMV H7 Camera Board
Author: Alasdair Allan