Capture screen using MSS and OpenCV for e.g. TensorFlow Object Detection API

In this video we will capture the desktop using MSS (Multiple Screen Shots) and OpenCV. With a little adaption you will also be able to store the recorded video on your hard drive.

Note: This tutorial utilizes Python 3.6.8 on Anaconda in a Windows 10 environment.

Other used libraries:

NumPy 1.16.3
OpenCV/cv2 3.4.2
win32api (pip install pypiwin32)
mss (pip install mss)

This approach will also include dynamic resizing of the screen capture window:

a - move capture left
d - move capture right
s - move capture down
w - move capture up
f - increase height
g - increase width
h - decrease height
j - decrease width

Python Code:

In the next videos I will go into details on how to utilize this for e.g. TensorFlow Object Detection API.

For this video, I used this traffic time lapse video:

Useful links:

[Jon-ss]

Hello! You can transmit this using a socket locally to another device.

[aldrinvelasco]

sir need help, , how you can combine your tensorflow detection in screen capture ?