TopolearnVideo is just a stream of images, and most real-time vision apps are a loop over VideoCapture frames.
cv2.VideoCapture(source) opens a video file, an image sequence pattern, a stream URL (RTSP, HTTP) or a camera by index (0 is the default webcam). Check cap.isOpened(), then loop:
while True:
ok, frame = cap.read()
if not ok:
breakread() returns (False, None) at the end of a file or when a camera drops, so always check ok. Each frame is a normal BGR array.
cap.get(cv2.CAP_PROP_FPS), CAP_PROP_FRAME_COUNT, CAP_PROP_FRAME_WIDTH and HEIGHT report properties; cap.set(...) requests camera settings, which the driver may ignore. Frame counts from containers are sometimes estimates. Call cap.release() when done.
cv2.VideoWriter(path, fourcc, fps, (width, height)) writes frames. The codec is a four-character code such as cv2.VideoWriter_fourcc(*"mp4v") for .mp4 or *"MJPG" for .avi. Every frame you write must exactly match the declared size and be 3-channel unless you pass isColor=False; mismatched frames are silently dropped, leaving an empty file. Available codecs depend on how OpenCV was built (FFmpeg, GStreamer, platform APIs).
For a live preview, cv2.imshow plus cv2.waitKey(1) in the loop gives about real-time display and lets you exit on a key.
Example
import cv2
import numpy as np
w, h, fps = 160, 120, 10
out = cv2.VideoWriter("demo.avi", cv2.VideoWriter_fourcc(*"MJPG"), fps, (w, h))
for i in range(30):
frame = np.zeros((h, w, 3), np.uint8)
cv2.circle(frame, (5 * i, 60), 10, (0, 255, 255), -1) # moving dot
out.write(frame)
out.release()
cap = cv2.VideoCapture("demo.avi")
print(cap.isOpened(), cap.get(cv2.CAP_PROP_FPS)) # True 10.0
n = 0
while True:
ok, frame = cap.read()
if not ok:
break
n += 1
cap.release()
print(n, frame is None) # 30 TrueRemember
- Loop
ok, frame = cap.read()and stop when ok is False - VideoCapture opens files, URLs and cameras by index
- VideoWriter needs a fourcc, fps and the exact (width, height) of every frame
- Release captures and writers when done
Watch out for
- Writing frames whose size differs from the VideoWriter's declared size, which produces an empty video without any error.