Leveraging WebGPU for Real-Time Machine Learning Visualizations in the Browser

This paper investigates the potential of WebGL and WebGPU for real-time scientific visualization in web browsers. By implementing comparable rendering tasks using both APIs, it evaluates their respective strengths in terms of implementation complexity, rendering performance, and scalability. WebGL, while accessible and widely supported, shows limitations in handling high-performance or parallelized tasks. WebGPU, despite its steeper learning curve, offers granular control and native compute-shader support, positioning it as a forward-looking alternative for computation-intensive web applications. Through controlled benchmarks and graphical comparisons, the study provides insights into selecting the appropriate API based on performance needs, development effort, and application complexity.