Vision AI-powered applications are exploding in terms of value and adoption across industries. They’re being developed both by sophisticated AI developers and those totally new to AI. Both types of developers are being challenged with more complex solution requirements and faster time to market. Building these vision AI solutions requires a scalable, distributed architecture and tools that increase developer productivity and can be accessible by those new to AI.
A cloud-native, microservice-based design is particularly well-suited to meet these requirements and provide huge benefits, such as scalability, serviceability, and being adaptable and deployable across a wide range of environments.
NVIDIA Metropolis Microservices
NVIDIA Metropolis Microservices is now open for early access, offering cloud-native, customizable building blocks to develop vision AI applications and services.
Metropolis Microservices offers abstracted, cloud-agnostic, enterprise-class building blocks that you can customize and integrate into your applications through APIs and industry-standard interfaces. You can deploy the microservices anywhere, from the edge to the cloud.
Metropolis Microservices enables enterprises and developers to accelerate the development of vision AI with ready-to-use reference applications, called AI workflows. Purpose-built microservices can be used as building blocks to develop bespoke solutions such as occupancy analytics, retail analytics, and more.
Reference applications and AI workflows that use these microservices offer a starting point for enterprises, systems integrators, and application developers to bring sophisticated vision AI services to market far faster and with less effort.
Early access microservices
The early access version includes the following microservices to help you build your customized vision AI application:
- Detection, Tracking, & Feature Extraction: Detects, tracks, and generates embedding for objects within each camera stream and handles multiple camera streams concurrently.
- Multi-Camera Tracking: Tracks movement of objects across multiple cameras.
- Behavior Analytics: Determines behaviors and anomalies in the movement of objects.
- Behavior Learning: Reveals and predicts behavior patterns through clustering and machine learning.
- Similarity Search: Performs high-throughput vector similarity searches using the Milvus database.
- Video Management & Storage: Discovers, connects, and manages cameras as well as managing video streaming and storage.
Early access AI workflows
The early access program also delivers the following AI workflows, which are prepackaged reference examples that use Metropolis Microservices, pretrained AI models, and the applications needed to jump-start your vision AI development:
- Retail Loss Prevention: Reduces misscans by giving existing kiosks visual recognition capabilities that can be adapted quickly to new products.
- Occupancy Analytics: Provides analytics on people’s movements such as space occupancy, heat maps, movement patterns, and so on. It can also be used for retail store analytics.
- Multi-camera Tracking: Tracks across cameras using people re-identification based on feature embedding and spatiotemporal association.
Speed up vision AI deployment
Unlock business insights faster than ever with existing video sensors across many physical spaces, including retail stores, airports, and urban roadways. Leverage a growing suite of reference applications to kickstart your development efforts. Metropolis Microservices makes it easy to build, test, and scale deployments from the edge to any cloud with enhanced resilience and security.
Metropolis Microservices and AI workflows are now available for early access. Apply now!
Check out the following upcoming GTC sessions on Metropolis Microservices: