RetroWinAug: A Lightweight Browser-Based Image Augmentation Tool with Privacy-Preserving Client-Side Processing

Image augmentation is a key step towards the quality and diversity of datasets to be used in applications such as in computer vision, machine learning, and deep learning. As traditional image augmentation tools often require software installations on the local machine, server-side processing or remove thcloud-based systems, issues relating to compatibility with the local system, computational resource demand, latency and data privacy are common. In this paper, RetroWinAug, a browser based lightweight image augmentation tool is introduced to overcome these constraints through client-side processing with keeping user's privacy and usability intact. With a retro inspired user-friendly interface, RetroWinAug enables users to do several important image augmentations such as grayscale conversion, rotation, flipping, sepia filtering, brightness variation, blurring, etc, in their browser. It displays the original and augmented image side by side, and allows to view the results instantly as visual feedback to improve workflow. Since RetroWinAug is implemented in HTML, CSS, JavaScript, and OpenCV.js, it all runs on the client side, thus no user's sensitive data, e.g. medical or proprietary images, need be transmitted to a server. It is cross platform and does not require any additional software, plugins or high-performance hardware and works on desktops, tablets and mobile devices smoothly. There are a number of applications that RetroWinAug is well suited for, including medical imaging, autonomous vehicles, satellite imagery, agriculture, academic research, educational environment, and so on. By combining all the usability, privacy protection and accessibility, RetroWinAug proposed a very efficient solution to the image pre-processing task across different domains.