Automated Image Classification with KNN and Vector Databases

This n8n workflow provides a robust solution for automated image classification, perfect for organizing and understanding large image datasets. It begins by generating dense vector embeddings for images using a sophisticated multimodal embedding model. These embeddings are then used to query a Qdrant vector database, retrieving the most similar image data points based on the KNN algorithm. A 'Majority Vote' logic is applied to the top results to determine the most probable classification, ensuring accuracy. The workflow intelligently handles potential ties by iteratively increasing the KNN search limit until a definitive classification is achieved or a predefined threshold is met, offering a scalable and reliable image categorization system.