Student

Chan Jia Liang

Supervisor

Ang Tan Fong

Collaborator

UM INNOVATIONS SDN. BHD.

The report describes the usage and benefits of Apache Kafka with ZooKeeper especially in managing a series of video frames. Real-time mask detection from live video stream aims to develop a real-time processing system that can accept live video streaming and identify the public who is not wearing masks from live video feeds using object detection algorithm.

As a face mask detection system grows, the application would require real-time processing on multiple sources of live video. However, video processing and analysis from multiple resources become slow when handling real-time video feeds. (Here, 2020) Since Kafka has a characteristic of high throughput with low latency, this makes big companies like Netflix and Spotify also implement Kafka framework into their system to ease the queueing of messages. To be able to handle and analyse a large stream of messages, Apache Kafka is chosen to act as the real-time stream processing framework in this project.

This project application uses Python programming language to develop the real-time mask detection system. Live video sources are collected and then converted into frames by Open Source Computer Vision (OpenCV) library. Then, the transmission of frames is handled and processed by the Kafka brokers which then stream to train and develop real-time detection on mask-wearing conditions using PyTorch with Faster R-CNN. The accomplishment of this project is to develop a real-time video processing system that is able to analyse, process and detect non-face mask wearers within a short period.