1. Why Corporate AI projects fail?
2. How AI Will Power the Next Wave of Healthcare Innovation?
3. Machine Learning by Using Regression Model
4. Top Data Science Platforms in 2021 Other than Kaggle
Running the demo with SSD
By default the script depthai_demo.py will load SSD
To run it just need to execute the script with no extra arguments, this will download the models and start the video by running the following:
Running the demo with YOLO V3
Running it with the following command:
python3 depthai_demo.py -cnn yolo-v3
The neural network frames per second with YOLO V3 (openVINO) was quite steady of 2 frames per second. Pretty fast considering how slow YOLO v3 can be.
I’ve benchmarked YOLOV3 in the past and while running from the compiled source of darknet, in this same machine it can take easily 18 seconds to analyse one single image.
Using YOLOV3 through openCv DNN module (which has been greatly optimised) it can take around 3 seconds.
So when I say it is pretty fast, I really mean it, 2 FPS is basically half a second per image, pretty good!
Testing the depth perception
Checking it with a far object, Z read 1.72 meters. I measured the distance with a tape and it was around 1.80 meters. So it was doing really good, also take into consideration I might not have followed the exact trajectory of the camera hence I could be off by some centimetres.
Checking the distance from the camera to my face, I measured 55 centimetres, and as you can see in the screenshot it was reading 62 centimetres, so great results overall.
Overall it was a good buy, would have like to avoid the surprise of the custom taxes but fair enough.
In terms of hardware it seems to be really good, the cameras are great and the casing itself is really good too, for what I could see so far it is performing really well, so looking forward to keep trying new things with it.
Did you find this article useful?
Share your comments and experience!, let us know what works and what did not work for you, do you have one? Or thinking on getting one?
Make sure to give this post some claps (50 or so is a good number!) if you enjoyed this post and want to see more.
And to keep up to date and read more articles on my adventures with the oak-d camera make sure to follow, until next time!