SEATTLE: Groundlight today announced the release of Counting Mode, making it easy for developers to reliably count how many items appear in an image, using a combination of advanced vision-language models, fast edge inference, and 24/7 live human supervision for ultimate reliability.

While many Video Management Systems (VMS) offer built-in analytics like people or vehicle counting, these features are based on pre-trained models, making them rigid and limited in scope. They’re typically designed for generic use cases and lack the flexibility to adapt to the unique needs of individual businesses. For example, a retail store might want to distinguish between the number of shoppers and employees in their stores, or alert security if somebody walks in with a roller-bag suitcase—tasks that go far beyond what most VMS analytics can handle out of the box. Without the ability to customize these visual insights, businesses are forced to either settle for incomplete data or invest heavily in complex, custom-built solutions.

AI-Powered Counting, No Machine Learning Expertise Required

Groundlight makes it easy to count the exact things you are looking to count, with no machine learning expertise or expensive setup to deploy. Businesses of all sizes can instantly and accurately count objects in any environment using natural language input. For most tasks, the ML model can be running within just a couple days. Users simply describe what they want to count in plain language, and Groundlight’s system does the rest—analyzing images from existing cameras to deliver fast, accurate results. With instant deployment, seamless API integration, and enterprise-grade accuracy, businesses can automate counting tasks and gain real-time insights without the hassle of traditional vision systems.

"Our goal is to make computer vision as easy as asking a question. With Counting Mode, businesses can quickly and accurately count objects in their environment—whether it's parts on an assembly line or products on a retail shelf—without needing a team of machine learning scientists."
— Leo Dirac, CTO at Groundlight AI

As an example, to get accurate footfall traffic data, you can count the number of people wearing your company uniform and exclude that data from total foot traffic. For developers looking to build this solution themselves, you can see the GitHub repo here for counting employees wearing your uniform. If the system is ever unsure of your requirements or of an image, it will escalate to Groundlight’s Cloud Labelers (humans in the loop), who will verify the answer in real-time. See this webinar recording where members of the Groundlight science team break this down further.

Key Features of Groundlight AI’s Counting Mode:

• Natural Language Interface: Count objects with human-readable queries.
• Fast and Accurate: Get real-time responses with high-quality visual outputs.
• 24/7 Human Annotation: Groundlight’s Cloud Labelers, or humans in the loop, take care of labeling in real-time.
• Simple SDK Support: Built into the Groundlight Python SDK for quick integration.

Counting Mode is now available to Groundlight users with a Business-level account, and documentation on how to create a counting application is available on the Python SDK Guide. Whether you’re building advanced AI-powered security applications, automating your manufacturing operations or getting custom retail analytics, Groundlight’s Counting Mode provides a fast and intuitive way to get object counts from images and streams.

To learn more, visit: code.groundlight.ai

Fonte: Business Wire