Скачать или смотреть How to build a custom GStreamer pipeline for the Raspberry PI 5 with Bookworm OS and RPi AI Camera.

#raspberrypi5 #raspberrypi #raspberrypiprojects

How to write a custom GStreamer pipeline then I use the cv2.VideoCapture() function to stream the video output to a monitor screen using the Raspberry Pi 5 and the Raspberry Pi AI (CSI) Camera .

There are some important concepts to keep in mind when writing a GStreamer pipeline.
At the top of the list when coding your pipeline data (buffers) flow from the source element (producer) pads to the sink (consumer) "pad" downstream. Each pad, be it a source (src) pad or sink(sink) pad must "negotiate" and either "approve" or "disapprove" the data being linked. If there is no data commonality between the src and sink pads then the data is blocked or dropped. This is why the gst-inspect-1.0 tool is important to use it on the elements and filters(plugins) to gain insight as to the capabilities of the pads.

This is my code snippet from the video:

#!/usr/bin/env python3

import sys
import gi
import cv2

gi.require_version("GLib", "2.0")
gi.require_version("GObject", "2.0")
gi.require_version("Gst", "1.0")

from gi.repository import Gst, GLib, GObject

Initialize GStreamer library
Gst.init(sys.argv[1:])

print(Gst.version_string())

pipeline = cv2.VideoCapture
("libcamerasrc camera name=/base/axi/pcie@1000120000/rp1/i2c@80000/imx500@1a ! \ videoconvert ! queue ! video/x-raw , pixel-aspect-ratio=(fraction)1/1, format=YUY2, \ width=640, height=640, num-buffers=10000000, framerate=30/1 ! queue ! videoconvert ! \ appsink ")

while pipeline.isOpened():
ret, frame = pipeline.read()
if ret:
cv2.imshow("Window", frame)
if cv2.waitKey(25) == ord('q'):
break
else:
break




pipeline.release()
cv2.destroyAllWindows()