[Paper Review] WearMask: Fast In-browser Face Mask Detection with Serverless Edge Computing for COVID-19

Zekun Wang, Pengwei Wang+3 more|arXiv (Cornell University)|Jan 4, 2021

Face recognition and analysisComputer Science18 references27 citations

TL;DR

WearMask delivers in-browser face mask detection using a serverless edge-computing framework that runs YOLO-Fastest via NCNN and WebAssembly in any web browser without installation.

ABSTRACT

The COVID-19 epidemic has been a significant healthcare challenge in the United States. According to the Centers for Disease Control and Prevention (CDC), COVID-19 infection is transmitted predominately by respiratory droplets generated when people breathe, talk, cough, or sneeze. Wearing a mask is the primary, effective, and convenient method of blocking 80% of all respiratory infections. Therefore, many face mask detection and monitoring systems have been developed to provide effective supervision for hospitals, airports, publication transportation, sports venues, and retail locations. However, the current commercial face mask detection systems are typically bundled with specific software or hardware, impeding public accessibility. In this paper, we propose an in-browser serverless edge-computing based face mask detection solution, called Web-based efficient AI recognition of masks (WearMask), which can be deployed on any common devices (e.g., cell phones, tablets, computers) that have internet connections using web browsers, without installing any software. The serverless edge-computing design minimizes the extra hardware costs (e.g., specific devices or cloud computing servers). The contribution of the proposed method is to provide a holistic edge-computing framework of integrating (1) deep learning models (YOLO), (2) high-performance neural network inference computing framework (NCNN), and (3) a stack-based virtual machine (WebAssembly). For end-users, our web-based solution has advantages of (1) serverless edge-computing design with minimal device limitation and privacy risk, (2) installation free deployment, (3) low computing requirements, and (4) high detection speed. Our WearMask application has been launched with public access at facemask-detection.com.

Motivation & Objective

• Motivate accessible mask detection to aid public health during COVID-19 by reducing installation barriers and hardware costs.
• Provide a device-agnostic, installation-free deployment that preserves user privacy by processing data locally.
• Develop a lightweight, fast inference pipeline suitable for edge devices without GPUs.
• Demonstrate integration of deep learning (YOLO), a high-performance inference framework (NCNN), and WebAssembly for browser-based deployment.

Proposed method

• Train a lightweight YOLO-Fastest detector (EfficientNet-lite encoder) on mixed real mask datasets (MAFA, WIDER FACE) and additional internet samples.
• Convert the PyTorch model to NCNN format and implement a C++ inference pipeline.
• Compile the inference pipeline to WebAssembly to run inside the browser via NCNN in a WASM module.
• Implement a serverless edge-computing design where video data is processed locally on the user’s device.
• Launch a publicly accessible in-browser demo at facemask-detection.com to showcase installation-free deployment.
• Benchmark detection performance (mAP@0.5) and inferencing speed on representative edge devices.

Experimental results

Research questions

• RQ1 Can in-browser, serverless edge computing support accurate and fast face mask detection without requiring installation or cloud resources?
• RQ2 What is the detection accuracy (mAP@0.5) and what FPS can be achieved on typical edge devices when running WearMask in a browser?
• RQ3 How does the WearMask pipeline integrate YOLO-Fastest, NCNN, and WebAssembly to enable device-agnostic deployment?
• RQ4 What are the privacy and deployment benefits of processing video locally rather than in the cloud?

Key findings

• The YOLO-Fastest model achieved a mean Average Precision at IoU=0.5 (mAP@0.5) of 0.89 after 120 training epochs.
• The WearMask framework is deployed as an in-browser application powered by NCNN and WebAssembly, enabling installation-free, device-agnostic edge inference.
• The system emphasizes low hardware requirements and privacy by processing video data locally on the end-user device, with a public demo website facemask-detection.com.
• The approach targets fast detection suitable for real-time scenarios across public facilities and small businesses, offering a cost-efficient alternative to cloud-based solutions.