AgCV: An Agentic framework for automating computer vision application

This paper proposed the Agentic Computer Vision (AgCV) framework designed to automate complex computer vision (CV) tasks through autonomous agents that communicate through Graphical User Interface (GUI). The AgCV framework leverages LangGraph, natural language processing, deep learning, and data science to build adaptive, user-driven CV pipelines. In the proposed AgCV each Agent works on a particular task ranging from object identification and classification to image segmentation. By incorporating Retrieval-Augmented Generation (RAG) and LangGraph, the AgCV enable fully automated pipelining through user interactions. The proposed Framework strategy reduces the need for technical expertise, allowing end-users to generate and configure CV operations using intuitive language commands. AgCV promotes accessibility, scalability, and flexibility of CV applications in different domains. The AgCV not only simplifies user interaction but also ensures that the system aligns with user expectations and needs.•The proposed system allows users to create and configure CV operations using simple natural language, making it accessible even to those with limited technical expertise.•The AgCV framework supports a wide range of CV tasks and can be easily adapted to different user needs and applications.