PY-L5-32 · Webcam Streams with OpenCV

Learning Goals

3 min

Open a webcam with VideoCapture.
Run the read → process → show → wait loop.
Quit cleanly and release the camera.
Run face detection live on the stream.

Warm-Up · A Video Is a Loop

5 min

open camera
while running:
    read a frame          (one image / NumPy array)
    process the frame     (anything from Lessons 30-31)
    show the frame
    check for quit key
release camera, close windows

Today's big idea

Real-time vision is just the single-image pipeline run in a tight loop on each frame. The only new things are opening the camera, the loop, and cleaning up afterwards. Note: this needs a local machine with a webcam — it won't run in Colab.

New Concept · The Capture Loop

14 min

The minimal loop

import cv2

cap = cv2.VideoCapture(0)        # 0 = default camera

while True:
    ok, frame = cap.read()       # ok=False if the camera fails
    if not ok:
        break

    cv2.imshow("Webcam", frame)  # show the frame in a window

    if cv2.waitKey(1) & 0xFF == ord("q"):   # press q to quit
        break

cap.release()
cv2.destroyAllWindows()

The pieces explained

VideoCapture(0)   open camera index 0
cap.read()        returns (success, frame); frame is a BGR array
cv2.imshow        draws the frame in a named window
cv2.waitKey(1)    wait 1ms for a key; required for the window to update
cap.release()     free the camera so other apps can use it

Process each frame

Insert any Lesson 30/31 operation between read and show:

ok, frame = cap.read()
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
edges = cv2.Canny(gray, 80, 160)
cv2.imshow("Edges", edges)

Overlay live text (e.g., FPS)

import time
prev = time.time()
# inside loop:
now = time.time(); fps = 1 / (now - prev); prev = now
cv2.putText(frame, f"{fps:.0f} FPS", (10, 30),
            cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 0), 2)

Always release

Forgetting cap.release() locks the camera so nothing else can use it (you'll have to restart). Wrap in try/finally if your processing might crash.

Worked Example · Live Face Boxes

12 min

# live_faces.py — webcam + real-time face detection
import cv2

face_cascade = cv2.CascadeClassifier(
    cv2.data.haarcascades + "haarcascade_frontalface_default.xml")

cap = cv2.VideoCapture(0)
try:
    while True:
        ok, frame = cap.read()
        if not ok:
            break

        gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
        faces = face_cascade.detectMultiScale(gray, 1.2, 5, minSize=(60, 60))

        for (x, y, w, h) in faces:
            cv2.rectangle(frame, (x, y), (x + w, y + h), (0, 255, 0), 2)
        cv2.putText(frame, f"faces: {len(faces)}", (10, 30),
                    cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 0), 2)

        cv2.imshow("Live Faces (press q)", frame)
        if cv2.waitKey(1) & 0xFF == ord("q"):
            break
finally:
    cap.release()
    cv2.destroyAllWindows()

Read the diff

The face-detection code from Lesson 31, now inside the capture loop — boxes follow you in real time. Two robustness touches: a higher scaleFactor (1.2) keeps it fast enough for live video, and the try/finally guarantees the camera is released even if something throws. That's the difference between a demo and a tool.

Try It Yourself

13 min

01 🟢 Mirror cam

Show your webcam flipped horizontally (selfie-mirror) with cv2.flip(frame, 1).

02 🟡 Snapshot key

Press s to save the current frame to a numbered file, q to quit.

Hint

key = cv2.waitKey(1) & 0xFF
if key == ord("s"):
    cv2.imwrite(f"snap_{count}.jpg", frame); count += 1
elif key == ord("q"):
    break

03 🔴 Live filter switcher

Press number keys to switch the live filter: 1=normal, 2=gray, 3=edges, 4=blur. Show the active mode as text overlay.

Mini-Challenge · Record a Clip

8 min

Use cv2.VideoWriter to record the processed webcam stream (e.g., with face boxes) to an .mp4 file while it plays. Stop recording on q.

Show the writer setup

fourcc = cv2.VideoWriter_fourcc(*"mp4v")
w = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
h = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
writer = cv2.VideoWriter("out.mp4", fourcc, 20.0, (w, h))
# inside loop, after processing:
writer.write(frame)
# after loop:
writer.release()

Non-negotiables: match the writer's frame size to the capture, release the writer too. Now you can save demos of your vision apps.

Recap

3 min

Webcam = VideoCapture(0) + a read/process/show/waitKey loop + release at the end. Each frame is a BGR array you can run any Lesson 30/31 operation on. waitKey(1) is mandatory for the window to refresh and read keys. Always release the camera (try/finally). Next: combine it into a smile detector.

Vocabulary Card

VideoCapture: OpenCV's camera/video reader; index 0 = default webcam.
frame: One image from the video stream — a BGR NumPy array.
waitKey: Waits N ms for a keypress; required to refresh the display window.
release: Frees the camera/writer so other programs can use it.

Homework

4 min

Build a live webcam app with: a mirror toggle, a filter switcher (normal/gray/edges), live face boxes, an on-screen FPS counter, snapshot on s, and quit on q. Clean release with try/finally.

open camera while running: read a frame (one image / NumPy array) process the frame (anything from Lessons 30-31) show the frame check for quit key release camera, close windows

import cv2 cap = cv2.VideoCapture(0) # 0 = default camera while True: ok, frame = cap.read() # ok=False if the camera fails if not ok: break cv2.imshow("Webcam", frame) # show the frame in a window if cv2.waitKey(1) & 0xFF == ord("q"): # press q to quit break cap.release() cv2.destroyAllWindows()

VideoCapture(0) open camera index 0 cap.read() returns (success, frame); frame is a BGR array cv2.imshow draws the frame in a named window cv2.waitKey(1) wait 1ms for a key; required for the window to update cap.release() free the camera so other apps can use it

import time prev = time.time() # inside loop: now = time.time(); fps = 1 / (now - prev); prev = now cv2.putText(frame, f"{fps:.0f} FPS", (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 0), 2)

# live_faces.py — webcam + real-time face detection import cv2 face_cascade = cv2.CascadeClassifier( cv2.data.haarcascades + "haarcascade_frontalface_default.xml") cap = cv2.VideoCapture(0) try: while True: ok, frame = cap.read() if not ok: break gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY) faces = face_cascade.detectMultiScale(gray, 1.2, 5, minSize=(60, 60)) for (x, y, w, h) in faces: cv2.rectangle(frame, (x, y), (x + w, y + h), (0, 255, 0), 2) cv2.putText(frame, f"faces: {len(faces)}", (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 0), 2) cv2.imshow("Live Faces (press q)", frame) if cv2.waitKey(1) & 0xFF == ord("q"): break finally: cap.release() cv2.destroyAllWindows()

fourcc = cv2.VideoWriter_fourcc(*"mp4v") w = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)) h = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)) writer = cv2.VideoWriter("out.mp4", fourcc, 20.0, (w, h)) # inside loop, after processing: writer.write(frame) # after loop: writer.release()