Client Infrastructure

Client-Provided Infrastructure for AI Inference

AVEKSHA and Bharat AI Vision models are software products. All hardware required for private or hybrid inference must be provided and maintained by the client or implementation partner.

Sizing depends on

Number of cameras

Camera resolution

Stream codec

Required FPS

Number of AI modules

Model complexity

Scene complexity

Required latency

Concurrent streams

Storage retention

Redundancy requirements

Deployment type

Indicative planning guidance

Evaluation environment

  • Approximately 1-5 selected cameras
  • Modern 8-core CPU or better
  • 16-32 GB RAM
  • Supported NVIDIA GPU with approximately 8-12 GB VRAM
  • 512 GB SSD or NVMe
  • Approved Ubuntu/Linux version

Pilot environment

  • Approximately 5-20 selected cameras
  • Modern 12-16 core CPU or better
  • 32-64 GB RAM
  • Supported NVIDIA GPU with approximately 16-24 GB VRAM
  • Minimum 1 TB NVMe recommended
  • Event storage based on retention requirements

Production environment

  • Final capacity determined after pilot and benchmarking
  • Enterprise server-grade CPU
  • 64 GB RAM or higher
  • One or more supported GPUs
  • Client-managed NVR/VMS/NAS/SAN
  • Redundancy based on operational requirements

Large deployment

  • Distributed inference nodes
  • Multiple client-provided GPU servers
  • Camera distribution across nodes
  • Central dashboard
  • Load balancing where required
  • Redundancy and failover where required
These specifications are indicative planning guidance only. Final hardware requirements must be confirmed through camera-stream, AI-model and inference-performance benchmarking.

Camera and Stream Requirements

Supported or assessable sources

  • IP cameras
  • RTSP streams
  • ONVIF-compatible cameras
  • HTTP video streams
  • Approved NVR/VMS streams
  • PTZ cameras
  • Thermal cameras
  • IR cameras
  • Night-vision cameras
  • Drone video streams
  • Recorded video files

Information required from the client

  • Camera brand
  • Camera model
  • Stream URL
  • Resolution
  • Codec
  • FPS
  • Main-stream details
  • Sub-stream details
  • Camera credentials
  • Camera location
  • Camera angle
  • Lighting conditions
  • Required AI module
  • Detection zone
  • Recording requirement
  • Retention requirement
Camera compatibility and AI performance are validated through technical assessment and pilot testing.
Bharat AI Vision Software

Benchmark before production sizing.

Validate streams, models and inference speed before final infrastructure decisions.

Start Pilot