We use Compute Module 3+ Lite in our Starter and Deluxe kits. But some of kits owners faced a problem - their StereoPi was not able to boot with installed micro SD. After digging out we found, that actually these users received kits with 8 GB eMMC equipped Compute Modules, but not Lite editions. Actually, these Compute Modules a bit more expensive then Lite. But they are unable to work with micro SD cards, as we described at the beginning of this video:
After detailed situation analysis and discussion with our kitting partner, we were able to find a reason of this wrong kitting. Looks like within a hundreds of Compute Modules 3+ Lite, used for assembling a kits, one wrong box with eMMC equipped CM3+ has been used. So we can expect about 30 non-Lite Compute Modules in our previous batch (1200 pcs).
If you’re one of these users, who got eMMC revision - you may use it as shown in a video mentioned. But if you definitely need Lite version - please contact us, and we’ll do CM3+ replacement for you!
In most of our tutorials we use Python language. You often hear that Python is too slow for computer vision, especially when it comes to single-board computers like Raspberry Pi. In this article, we decided to measure the actual speed difference and find the performance ‘bottleneck’. The approach is very simple. We have a series of small Python programs that allow you to go through all the stages from the first launch of the stereo camera and its calibration to building a depth map from real-time video (and a 2D space map in a mode that emulates the operation of a 2D lidar). We ported all this code to C++, and we compare performance at each stage. You can read full article here in our blog.
We’re in a process of moving all articles from our blog on Medium to our own site. So you can find all articles and news in a one place here: https://stereopi.com/blog If you find some issues with our new blog - please don’t hesitate to inform us on our forum.