OpenCV3 renderer is too slow due to cv2.waitKey(1). If you want to more performance, you should use OpenCV4+ or 'pyglview' package.
This package is supported fastest OpenGL direct viewer and OpenCV renderer both. If your environment was not supported OpenGL, it will be switched to CPU renderer(OpenCV) automatically and also available remote desktop(Xserver) like VNC.
pyglview is useful library instead of OpenCV imshow/waitKey API.
curl -sL http://install.aieater.com/setup_rocm | bash -
curl -sL http://install.aieater.com/setup_nvidia_with_cuda10 | sudo bash -
Ubuntu16
sudo apt update
sudo apt install -y ubuntu-desktop
For remote desktop.
sudo apt install -y gnome-panel gnome-settings-daemon metacity nautilus gnome-terminal vnc4server
Also see xstartup template script => http://install.aieater.com/xstartup
Install OpenGL native packages, (Ubuntu16/18)
# Full OpenGL packages.
sudo apt install -y build-essential
sudo apt install -y libgtkglext1 libgtkglext1-dev
sudo apt install -y libgl1-mesa-dev libglu1-mesa-dev mesa-utils
sudo apt install -y freeglut3-dev libglew1.10 libglew-dev libgl1-mesa-glx libxmu-dev
sudo apt install -y libglew-dev libsdl2-dev libsdl2-image-dev libglm-dev libfreetype6-dev
Install python packages (Linux/MacOSX/Windows)
pip3 install PyOpenGL PyOpenGL_accelerate
| Version | Library | installation |
|---|---|---|
| v3.x/v4.x | OpenCV | pip3 install opencv-python |
| v1.1x.x | numpy | pip3 install numpy |
| v3.1.x | PyOpenGL | pip3 install PyOpenGL |
| v3.7.x | configparser | pip3 install configparser |
pip3 install pyglview
import cv2
import pyglview
viewer = pyglview.Viewer()
cap = cv2.VideoCapture(os.path.join(os.path.expanduser('~'),"test.mp4"))
def loop():
check,frame = cap.read()
frame = cv2.cvtColor(frame,cv2.COLOR_BGR2RGB)
if check:
viewer.set_image(frame)
viewer.set_loop(loop)
viewer.start()
If you got something warning message, it is not using GPU.
Use GPU directly
This message is successful to use GPU.
@WARNING: No display.
No display message meaning is there is no available display. This message will be appeared, when you have executed program in ssh console.
@WARNING: GPU or physical display is not available. Use CPU renderer.
This message will be appeared, when there was no GPU or GPU driver, or remote logged in like VNC. You have to make sure GPU driver and PyOpenGL packages and use physical display.
If you want to more performance and non-blocking API for camera and video, acapture package is very useful.
import cv2
import acapture
import pyglview
viewer = pyglview.Viewer()
cap = acapture.open(0) # Camera 0, /dev/video0
def loop():
check,frame = cap.read() # non-blocking
frame = cv2.cvtColor(frame,cv2.COLOR_BGR2RGB)
if check:
viewer.set_image(frame)
viewer.set_loop(loop)
viewer.start()
Logicool C922 1280x720(HD) is supported 60FPS. This camera device and OpenGL direct renderer is best practice. Logicool BRIO 90FPS camera is also good!, but little bit expensive.
This project is licensed under the MIT License - see the LICENSE file for details
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent