Yes, there were no videos out there comparing these kind of cameras until now, thanks for this! 🥰 I'm not a vision specialist, I built a small custom driving ROS robot using Asus Xtion Pro Live together with a 360° Lidar some years ago. This camera is pretty old but still buyable somwhere. I was looking for a new one in the lower price segment. I tried an OAK cam but sent it back immeadialty because for me it was much worse than the ASUS for my purpose. Then I tried a realsense D435i which I still have. This camera is ok for me but I'm disapoined that it is not much better than the Asus Xtion Pro Live, at least for my use case. For my purpose, the navigation, object avoidence, the Asus Cam is still the best because it is so basic but has still a good point cloud field which is barely flickering.
I love seeing the side-by-side test. Ive always been impressed by the data quality of the Azure Kinect. Microsoft discontinued it, but they handed it off to Orbbec which now makes an essentially identical sensor, as well as a few variants of it.
I like when someone screencasts as he codes like Aladdin Persson or Andrey Karpaty making yolo or gpt transformer from scratch. You could show how to run yolov5 or other networks on rknn2. Another thing I like to know is Intels Neural Accelerator (part of N100 and N300) and AI Boost (Core Ultra). How do they compare with each other and rockchip NPUs?
But... This video is about cameras!...:) "yolov5 or other networks on rknn2" - ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-pN8mKZ5wpdQ.html ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-gYIMYd2x52A.html N100 - maybe I will test it. Don't know yet. "AI Boost" is not this channel's main target (not an edge board). But maybe if I have it, I will test it.
i think kinect suffers the most with multipath due to the nature of the sensor, sometimes V shaped corners will appear rounded, but otherwise it has the best performance out of any other similar depth sensors, orbbec licensed their technology and they have a slightly updated one now that's pretty good, also i think analog device has a few iDoF modules that's more integratable for hardware projects that perform very well
Would you consider testing these cameras underwater to see how good of a depth each one can achieve? By the way, very nice video. Greetings from Chile!
Thank you! I think that stereo cameras should be better than active illumination cameras. Never have such tasks. Sadly, I don't have plans to test this shortly. What task for this do you have? Fish size measurement?
Never had such tasks. I think that there was some solution for RealSense: dev.intelrealsense.com/docs/multiple-depth-cameras-configuration www.intelrealsense.com/how-to-multiple-camera-setup-with-ros/?_ga=2.45937892.153457549.1711046876-272933498.1710111396 But never tested this myself.
hi there i have oak d pro and use in unity. i see example for point cloud vfx but the distance only show up 1.5 Meters, how to make more far the distance? thanks in advance
I don't think it will be better. The same physics behind the same algorithms. It may be slightly better because of the different distances between lenses.
@@Calm_And_Peace_ What do you mean by outdoor? Algorithms are the same:) But if you speak about box protection, I think that you need additional cover.
