(Flood image from Indian Navy)
A Vision Language Model (VLM) is a particular type of Large Multimodal Model (LMM).
“Vision language models are broadly defined as multimodal models that can learn from images and text. They are a type of generative models that take image and text inputs, and generate text outputs…. There’s a lot of diversity within the existing set of large vision language models, the data they were trained on, how they encode images, and, thus, their capabilities.”
“VISense (is) a groundbreaking addition to its VISuite platform that redefines real-time video analytics with Vision Language Models (VLMs). VISense represents a major advancement in Generative AI integration, using VLMs to achieve detailed scene understanding and contextual insights empowering operators to make informed decisions promptly….
“VISense allows users to ask questions like, “Let me know when something unusual is happening in any camera view” and receive a detailed response describing the unusual aspect of the captured behaviour. For instance, it might respond, “Yes, there is a flood; water levels are rising in the northern section, and several vehicles are stranded, causing heavy traffic congestion,” providing actionable insights that enable quick decisions.”
Bredemarket 