I'm working on an application in python using opencv for use in athletic coaching. I'd like it to run at 60fps, but I can't seem to get it to that speed. I've determined that the part that is slowing everything down is the capture process; frames aren't getting from the camera to the cpu fast enough.
Now, I know it is possible for the camera to run at 60fps, because when I open up the same camera in guvcview (at the same resolution), I can get a full 60fps.
I don't want to share my entire project, but here's the part that manages frame capture: https://pastebin.com/nnx32rQK
Any help would be welcome!
Related
I've been searching for quite some time now, but the only example I've seen where camera motion keyframes (movements) can be extracted from a video is in blender. I'm looking for a script, library, colab notebook, or any recommendation to be able input a video (for example, getting the camera keyframe movements from a video of a drone flying around) and exporting it to any type of file. The main purpose of this is double-sided: I want to be able to take the camera keyframe movements from the drone video and input it into an AI Art Animation and I want to be able to learn and add this python tool or library into a colab notebook of AI tools I've been slowly building over the months. Any help from you masterful wizards will be appreciated. Thanks!
I'm a "hacker", "novice-intermediate" level python coder, meaning I can read code and understand most basic and some advanced type coding. I'm able to read code quite well and manipulate it to my preferences, but I just don't even know where to start when it comes to the request I'm asking, ergo my post to Stack.
I am trying to stream live video from an external camera using cv2. I was able to write the simple code to stitch the frames and stream it. But am struggling to find how to change the camera.
I tried to run it after disabling the main webcam from the task manager, but it still did not work.
So, if anyone can help me with some clue regarding the same, that would be a great help.
Cameras are numbered for Windows. You can try a few indeces and check which camera index belongs to the camera you want.
capture = cv2.VideoCapture(index)
I am trying some video processing exercises, and wondering if there is a way to display multiple video streams into one window a la pyplot.subplot command for the images.
I've tried using subplot syntax, but using it freezes the program, so any alternative source would be much appreciated.
It is possible to display multiple videos(e.g 2-videos can be displayed for dual core processor) at a time for that you have use OpenMP. It is possible if your using opencv. And it is pretty easy Only you have to enable OpenMp in your property sheet.
I have been working with some drones and robotics projects using arduino and python. There was a kickstarter project for a neat little hex copter, that hasn't been managed well.
I was lucky, i got my copter and then some time later after some frustrated email exchanges, i finally recieved the camera as well. To this day, their forum has people still complaining. Their maker forum is now down and their wiki hasn't been updated with any specifics on the camera.
http://www.flexbot.cc/wiki/index.php?title=Main_Page#Hardware
Their app to accompany the drone still doesn't support the camera module. Not that it'd matter, as their code isn't very well documented or annotated.
https://github.com/HexAirbot
There are some tips on switching the camera on the comments page of their kickstarter campaign.
https://www.kickstarter.com/projects/1387330585/hex-a-copter-that-anyone-can-fly/posts/1093716
So, sob story over, i'm stuck with this neat little wifi camera that i am unsure on how to connect to. I know how to switch it on and it does have a micro-usb port on it.
What library in Python could i use to stream an image from this camera given that it is a wifi camera. If i wanted the video stream as a numpy matrix.
I need to interface with the camera, so i can connect and disconnect.
Then, be able to read images frame by frame with ffmpeg. I have some python modules that can detect and read from a camera, but how can my code ensure that the camera is connected?
Totally stuck. Any help would be appreciated.
Considering you are building for the android platform, you will more than likely need to use some sort of java/python driver/interface, unless you just use java.
Here is an article on java/python, and using python from within java.
Using Python from within Java
Background
I'm attempting to craft a simple video playback script for a small cinema that automates the playing of videos and control of the projector, sound and lighting systems. I have two video outputs, one goes to a monitor in the projection booth, and the other directly to the projector. I desire to play video (and only video) fullscreen to the projector while putting controls and a small (~1/4 screen) preview on the monitor. This will allow the projectionist to view the video being output and control the playback from the monitor in the booth while all the audience ever sees is the video output.
Problem
I am currently using Python to control VLC player (with libvlc Python bindings) to playback videos. I have everything working fine except that I can't figure out how to get a preview (direct copy) of the video being played fullscreen on the projector output into my GUI.
I have tried using the clone filter, but I cant get the cloned window to automagically appear full screen nor in my GUI. The clone filter seems like the logical choice but it seems to be VERY inflexible when it comes to specifying destination screens, fullscreen, etc. I must be able to open video windows full screen on the projector monitor. Professionalism is key and it would look bad if the projectionist had to drag a window over and double click on it when the movie started.
Currently Using:
Debian Linux
Python 2.7
wxPython
libvlc
I would like to continue using Python as I already have the code for controlling the projector, sound processor, lighting and curtain written and tested. I chose VLC because it really seems bulletproof when it comes to video playback but am not committed to it's continued use. I also chose wxWidgets for my GUI as a result of past experience but I am not stuck on that either.
This describes the direct solution and does not concentrate on any alternative or the overall design of your application.
As Your Application and VLC media player are separate processes, you will not be able to get what you want directly because there is no "shared memory" between those 2 applications. The best shot to "copy" the decoded frames from VLC will be to e.g. send a RAW Video .mts stream (ts is usually used for this kind of usecase) and send e.g. to udp://localhost:1234.
In your application, you will need to be able to receive the ts stream, "decode" it and display at the spot of interest.
For start, i would try if you are able to do this using 2 vlc players that you control manually. When you achieved that the first VLC streams to udp and outputs on the main display at the same time, and the other VLC player receives and plays the udp stream you can go on:
Find a player library that you can use directly in your wxpython application and check if it can receive the udp stream as well E.g.
https://wxpython.org/Phoenix/docs/html/wx.media.MediaCtrl.html
This player lib for example requires gstreamer as a base.
As a result, main display and the picture in your applicatoin might have a latency of some seconds. To come around this latency, the best way that i currently know is using WebRTC but this is a lot more complex setup than the above.
https://www.sipwise.org/news/technical/tv-over-webrt/
Sure in case you do some "encoding" for WebRTC or even for UDP, you would need to utilize some hardware encoder, e.g. Nvidia NVENC in order to be able to guarantee the needed resources are always there.