VAS SDK

Enterprises, system integrators (SIs), OEMs, and developers can take control with Megh’s Open Analytics VAS SDK: A simplified and supported platform for custom, high-performance intelligent video analytics solutions.
VAS SDK allows developers to optimize video (and data) analytics pipelines and integrate highly customizable AI into applications and products.
With AI+ acceleration using GPUs, FPGAs, and SOCs, VAS SDK offers the highest performance, most scalability, greatest flexibility, and best vendor choice for system integrators and independent software vendors.
Features
VAS SDK is a platform optimized for real-world complexity and performance.
Fully customizable pipeline
- End-to-end configuration of the analytics pipeline (ingestion, transformation, inferencing, filtering, and visualization)
- AI Models: Swap or retrain/redeploy AI models for single- or multiple-stage inferencing
- Filtering libraries: Configure parameters or deploy new libraries to implement business rules
- Extensible pipeline: Ability to include additional pipeline elements as needed
- APIs for easy integration with business, messaging, or security systems (VMS, BI/visualization, and more)
Real-time system
- AI optimized to deliver real-time performance with lowest latency
- Pre-tested hardware configurations to support performance requirements
- Easily change compute platform on the fly with one line of code
- DLE inference accelerator on FPGAs delivers best-in-class throughput
Cross-platform support
- Easy migration across compute platforms (CPU, GPU, FPGA, SOC)
- Multiple deployment models (edge, on-prem, private and public cloud, hybrid)
- Integrates easily with any IP camera or sensor
Open vendor choice
- CPU (Intel, AMD, ARM)
- GPUs (Nvidia), FPGAs (Intel, Xilinx), and SOCs (Qualcomm)
- Cloud hosts (AWS, Azure, Google)
- OEM platforms (Dell, HPE, Lenovo, IBM)
Intelligent video analytics
VAS SDK enables development of end-to-end intelligent video analytics solutions with the following Megh products:
VAS Suite— Deploy market-ready solutions that create actionable insights to reduce risks and improve operational efficiencies.
VAS Portal— Manage multiple VAS Suite instances via any browser with our easy-to-use dashboard.
For help with creating or implementing video analytics for your custom use case, contact us.
Specifications and implementation
- Languages supported: Python
- AI model frameworks: Pytorch and TensorFlow (TF-Lite)
- Platforms: CPU, GPU, FPGA, SOC
- Vendors: Intel, AMD, ARM, Xilinx, NVIDIA, Qualcomm
- Cloud infrastructure supported: AWS, Microsoft Azure, Google Cloud
For more technical details, learn about The Megh Platform.
SDK extensions
Enhance VAS SDK with our SDK extensions:
- DLE SDK: Train, build, and deploy custom models with Pytorch, TF-Lite, and other frameworks on FPGAs for AI deep learning inferencing
- AFU SDK (coming soon): Develop custom FPGA accelerator function units (AFUs) for analytics using RTL/HLS for deployment on scale-out FPGA platforms
For more information about DLE SDK, contact us.
Why VAS SDK?
VAS SDK is an Open Analytics solution, designed to enable flexible, scalable, high-performance, end-to-end integration into your business operations, applications, and products.
Comprehensive control
- Customize the AI/video analytics pipeline for any camera/sensor stream
- Optimize for accuracy, TCO, or performance based on use case
- Edge-to-cloud deployment options for performance, cost, and regulatory adherence
Safe scalability
- Start with CPU and scale to GPU/FPGA as needed
- Change compute platform and/or deployment type on the fly
- Stability for business users while scaling (no need to recode app layer)
Easy AI adoption
- Utilities provided to help report data and optimize analytics
- Concierge Support for custom integration or model development
Optimized TCO
- Right-size hardware for simple-to-complex use cases, based on pre-tested specs
- Lowest TCO for any performance target
Enables operational reliability
- Intelligent video analytics that actually deliver in complex environments and use cases


Product brief
Technical papers
This 11-page Dell-EMC technical white paper evaluates a video analytics pipeline with deep learning inference using Megh’s VAS and Intel Programmable Acceleration Card (PAC) FPGAs on Dell EMC PowerEdge R740/R740xd servers.
This 3-page HPE technical white paper discusses how customers can achieve dramatically increased performance and reduced TCO by running the Megh Computing real-time streaming video analytics platform on the HPE ProLiant DL380 Gen10 server with Intel Arria 10 GX FPGAs compared to a configuration without FPGAs.
Intel partnership
In the following video, see how Intel and Megh have partnered in developing low-latency analytics using AI with Intel FPGAs.
