San Francisco-based OpenMV has a vision to become the “Arduino of Machine Vision.” After the success of its machine vision camera, the Cam M7, the company is back with the new and improved Cam H7, a next-generation cam that follows the success of OpenMV Cam M7, which has sold more than 10,000 units. Details of the camera were launched on the crowdfunding platform Kickstarter on September 17, with a campaign for $50,000.
The Cam H7is a 100 percent open-source MicroPython-powered machine vision camera designed for low-power real-time applications that works similarly to Arduino. The camera can be programmed in Python 3. Using Python instead of C gives users less than 1-second compile/load/execute times to quickly work on projects. However, users still have the option to program directly in C.
The new H7, which is twice as powerful as the Cam M7, has features such as removable camera modules for global shutter and thermal vision support, as well as a standard rolling-shutter color camera sensor. Users can run machine vision algorithms on what the OpenMV Cam sees, enabling them to track colors, detect faces in seconds, and control I/O pins in the real world.
Multi-blob color tracking can detect an unlimited number of blobs per frame, which can be found by segmenting the image with up to 16 color thresholds. Color threshold selection is simplified through the Threshold Editor tool in OpenMV IDE.
The Cam H7 can track April tags, decode QR codes, read a wide range of bar codes, and parse data matrices twice as fast as the Cam M7. It also includes FLIR Lepton 1/2/3 thermal camera sensor support, which is exclusive to the Cam H7 and allows users to run all algorithms on thermal images.
The MT9VO34 camera sensor module, a grayscale-only global shutter sensor with on-demand trigger support, provides support for those who require precision machine vision support. . The sensor can run at 80 FPS at VGA/QVGA, 200 FPS at QQVGA, and 400 FPS at QQQVGA. The Cam H7 also supports loadable neural networks, and it is capable of running simple CNN's like LENET-6 for number detection.
OpenMV created its machine vision cameras to “bring machine vision algorithms closer to makers and hobbyists” and reduce time-consuming algorithm work so that users can focus on the creative process instead.
As of writing, OpenMV has received $95,596 from 646 backers, which is $45,596 above its $50,000 goal. After the Kickstarter campaign, OpenMV plans to focus on training networks to conduct people detection for the Cam H7.