So I am interested in using my raspi as a robot and I want to try to get a video stream from a camera attached to the raspi. I found a tutorial on how to get the video streamed over LAN using python code (https://randomnerdtutorials.com/video-streaming-with-raspberry-pi-camera/). I don't understand how it works much but I'm happy with it as it has little delay compared to the VLC method of streaming I've seen suggested to stream. My main question is would I be able to use this video stream in something like OpenCv and then use the video for facial recognition. I'm not super familiar with python (although I am learning) so if I have a fundamental misunderstanding just let me know. Thanks in advance!
Ok, the answer to the question is it possible to do that is yes, if you have a video stream you can run object detection algorithms on the stream.
I'm sure you could also do this with tensorflow, but I haven't done anything with tensroflow beyond basic object detection following this guide.
Python is the common language for a lot of machine learning, but you're not necessarily tied to python.
If you want more information on how to do it after some research, it's probably more of a question for one of the other SE sites since it's not raspberry pi related if you're doing it on another computer. And you probably do since performance on the pi is quite low.
PiCamera has a usefull library for this. Full camera functionality is an option, from varying light level, exposure, previews, network stream piggybacking, the works. Me personally: i use the pi zero-w h inside my bots, and remote in with VNC. Using preview windows (which display to the 'desktop') while remoted in i have a nearly streamline laggless video stream because of the way vnc remotes in and the speeds processed via thst method(most network streams are sorely delayed due to compression and decompression times with lags upwards of seconds, and that can be catastrophic for remote droids) at most ive noted a 5 ms delay and that was using my phone and its 4g network from a state away(actually ran that test just earlier this week while away from home for work) which can be easily compensated for by running the bot at reduced speeds