BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

How Lumeo Simplifies Video Analytics With A Low/No Code Vision AI Platform

Following

Lumeo, a Bay Area-based startup, is building a platform to derive rich insights from existing surveillance cameras and CCTVs without the need to deal with complex AI pipelines and computer vision technology stack. Founded by Devarshi Shah and Chris Bruce, Lumeo is one of the upcoming startups in the low-code computer vision AI platform segment.

Vision AI represents one of the complex domains of artificial intelligence. Enterprises building computer vision solutions will have to invest in teams that are well-versed in deep learning, convolutional neural networks, MLOps, edge deployments, and business intelligence. To accomplish the end goal of integrating vision AI outcomes with existing analytics, the teams must perform plumbing to connect various existing and new components, which is laborious and expensive.

Lumeo aims to reduce the toil involved in building vision AI solutions. It simplifies tasks such as camera management, ingesting video feeds, processing, storage, transfer, deploying AI models, and correlating the inference results with existing KPIs of a business intelligence system.

Customers can bring any existing camera, video stream, media file, video management system and NVR to Lumeo with minimal effort. The cameras may be connected via RTSP, HLS streams, ONVIF, or even USB. Once the video source is connected to the platform, they can choose from a wide range of ready-to-use, curated vision AI models to perform inference. The output from these models can be visualized within the platform or extended to third-party destinations through extensible connectors.

Lumeo comes with a visual canvas that’s appealing to non-developers and business decision-makers. The canvas enables users to build a pipeline through a simple drag-and-drop interface. The tool allows custom code to extend the use case and scenarios for power users and technical audiences. It’s easy to prototype a solution and graduate it to production with minimal changes.

When I first saw the tool, it reminded me of Visio, a popular tool from Microsoft to diagram various technical architectures. The key difference is that, behind the scenes, Lumeo also generates the code and configuration needed for inference, deployment, and integration.

Since most of the cameras are deployed in remote locations, Lumeo also supports edge deployment. Customers can run a single command to connect their edge computing environment to the Lumeo cloud. The models can be deployed on powerful servers with NVIDIA GPUs or purpose-built, single-board computers like NVIDIA Jetson Nano. Lumeo supports both stand-alone deployments or Kubernetes-based distributed deployments. The process can be automated via the REST API exposed by the platform.

Lumeo has a forever-free starter tier for developers to use with their own compute devices running at the edge or in the public cloud. Enterprises can sign up for consumption-based pricing starting at $10 / stream/month for live streams or $100 for 100k video mins/month for clips.

The startup is partnering with silicon manufacturers such as Intel and NVIDIA and public cloud providers, including Microsoft Azure, AWS, and Google Cloud.

MORE FROM FORBESGoogle Adds A No-Code Computer Vision Platform To Vertex AI

The no-code/low-code development environments are now extended to artificial intelligence. Last year, Google launched the Vertex Vision AI platform - a no-code tool to run vision AI models in the cloud. I expect other vendors to jump the bandwagon in the coming months.

Follow me on Twitter or LinkedInCheck out my website