← Return to Portfolio

By Rommel Sharma · LinkedIn

Objective

This project aims to build a production-ready, real-time face recognition system. The focus is on achieving sub-100ms latency via WebRTC streaming while maintaining high-accuracy detection and embedding matching.

Core Technology Stack

InsightFace (buffalo_l) Gradio WebRTC ONNX Runtime Python HuggingFace Spaces

The system utilizes the InsightFace buffalo_l model pack (SCRFD detector + ResNet-50 ArcFace). Real-time performance is achieved using Gradio 5.x with FastRTC, removing standard HTTP polling latency.

ML Foundations

The architecture leverages 512-dimensional normalized embeddings. By applying ArcFace (Additive Angular Margin Loss), the model ensures highly discriminative embeddings, even in variable lighting. Inference is accelerated via ONNX Runtime for efficient cross-platform execution.

Development Roadmap

Note: This project is currently under active development.

← Return to Portfolio

Let's Connect

I am always open to discussing new challenges in the AI and Machine Learning space. Whether you are exploring how these patterns can be adapted for your specific domain, have questions about the architectural choices detailed above, or are looking to collaborate on impactful technology projects that help the society, I would love to hear from you.

Connect on LinkedIn