Experiment 6: Live Video Streaming

Edge AI / Computer Vision

Streaming Webcam Frames

In real-world Edge AI applications (like self-driving cars or security systems), we stream video frames to cloud servers for analysis. Here, we'll use OpenCV and Kafka to stream your webcam!

flag Aim

To capture live video frames using a webcam, compress them, serialize them into bytes, and stream them over a Kafka topic to a remote consumer that reconstructs and displays the video.

data_object Theory: OpenCV + Kafka

Video is just a sequence of image arrays. We capture an image, encode it to JPEG (to save bandwidth), and send the raw bytes over Kafka.

Producer (Capture & Encode)

import cv2

camera = cv2.VideoCapture(0)
ret, frame = camera.read()
# Compress frame to JPEG
_, buffer = cv2.imencode('.jpg', frame)
producer.send('video-stream', buffer.tobytes())

Consumer (Decode & Render)

import cv2, numpy as np

for msg in consumer:
    # Convert bytes back to Image
    nparr = np.frombuffer(msg.value, np.uint8)
    frame = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
    cv2.imshow('Video', frame)

checklist Mission Milestones

0 / 4 Done

radio_button_unchecked 1. Start the Kafka Broker

radio_button_unchecked 2. Run the Camera Producer

radio_button_unchecked 3. Run the Video Consumer

radio_button_unchecked 4. Stream at 2 FPS (Hint: change time.sleep(0.1) to 0.5)

Terminal 1 — Broker (192.168.1.10)

Offline

KRaft Broker — Ready to initialize.

1
2
3
4
5
6
7
8
9
10
11
12
13
14

terminal Console Output

Awaiting execution...

visibility Video Feeds

PRODUCER CAM

CONSUMER STREAM

FPS: 0 BW: 0 KB/s