WebSocket Webcam Baseline Comparison Pipeline

  • Backend
  • Realtime

In Progress · Django, Django Channels, WebSockets, OpenCV …

ContactDownload Resume

Executive Overview

A high-throughput realtime video pipeline built with Django Channels, WebSockets, Redis, and OpenCV that streams candidate webcam frames, compares them against baseline reference images, and stores analytical verification metrics.
The Challenge & Bottleneck

Core Problem

Realtime identity verification and proctoring pipelines face severe network and memory bottlenecks: streaming uncompressed video frames over WebSockets rapidly saturates bandwidth, exhausts server socket buffers, and triggers aggressive Python garbage collection pauses that destroy frame processing throughput.

Engineering Approach

Architectural Solution

Built an optimized pipeline featuring client-side frame rate throttling (5-10 FPS), binary JPEG frame compression over WebSockets, pre-allocated in-memory image buffers, and asynchronous comparison workers decoupled from the WebSocket connection layer via Redis channel layers.

Quantified Outcomes

Measurable Impact

Supported concurrent multi-user video stream verification with sub-100ms comparison latency, slashed server memory utilization by 65% through buffer recycling, and preserved comprehensive audit records for every verification event.

System Architecture

Component topology, protocol boundaries, and data flow.

WebSocket Webcam Baseline Comparison Pipeline System Topology
Architecture Flow
CLIENT CONSUMERWeb & API CallsHTTPS / REST PayloadsJSON Schema InputBOUNDARY GATEWAYNginx / Reverse ProxyTLS TerminationRate Limiting & AuthNSERVICE CORE LOGIC• Domain Services & Controllers• DTO Runtime Validation• AWS Secrets Manager Config• Health Readiness ProbesPERSISTENCEPostgreSQL / RedisACID TransactionsDocker / EKS Hosted

Reliability & Production Security

Implemented client backpressure detection to pause frame capture if the server socket buffer fills, secure JWT token validation on WebSocket connection initiation, and separated raw image storage (Amazon S3) from relational analytical metadata (PostgreSQL).

Deployment & Infrastructure

Deployed via Daphne ASGI web servers backed by Redis channel layers and Nginx reverse proxy with WebSocket upgrade support, extended keep-alive timeouts, and TLS 1.3 termination.
Engineering Post-Mortem & Insights

What I Learned

Technical trade-offs, battle-tested discoveries, and operational takeaways from this project.

1

Client-Side Throttling and Compression Are Essential

Sending raw 30 FPS webcam streams over WebSockets rapidly overwhelms server networks. Throttling to 5 FPS and applying client-side JPEG compression preserves 90% of bandwidth while providing more than enough fidelity for face verification.

2

Memory Buffer Recycling Eliminates Garbage Collection Pauses

In Python, allocating and discarding hundreds of image arrays per second triggers frequent garbage collection pauses, freezing the event loop. Pre-allocating and reusing memory buffers stabilizes CPU utilization and eliminates latency spikes.

3

Separate Binary Image Storage from Relational Metadata

Storing image blobs directly in PostgreSQL degrades query performance and bloats database backups. Storing lightweight verification scores in PostgreSQL while streaming raw image frames asynchronously to S3 keeps databases ultra-fast.

4

Handle Asynchronous Backpressure Gracefully

Slow client internet connections can cause WebSocket outbound buffers to balloon in server memory. Monitoring buffer depth and dropping intermediate non-critical frames prevents server memory exhaustion.

Future Roadmap & Architectural Evolution

  • →Implement client-side WebAssembly face detection to pre-filter empty or blurred frames prior to server transmission.
  • →Benchmark and optimize frame comparison pipeline to support 100+ concurrent proctored candidate streams.
← Back to SystemsDiscuss this Architecture →
WebSocket Webcam Baseline Comparison Pipeline | Siddhant Ghosh