Crop Anomaly Detection and Classification Workflow

This n8n workflow is designed for detecting anomalies in crop images by leveraging vector embeddings and similarity searches within a Qdrant collection. It automates the process from inputting an image URL to determining whether the crop depicted is within the known dataset or anomalous. The workflow begins with an input trigger where users submit an image URL. This URL is then embedded into a vector representation using the VoyageAI API. The resulting vector is compared against pre-established medoids (central points) of crop classes stored in Qdrant, a vector similarity database. Using a custom Python script, the workflow assesses similarity scores to identify potential crop identities or anomalies. Additional nodes fetch detailed crop class information and count the number of crop classes in the collection. The entire process supports agricultural research, quality control, and crop monitoring by automatically identifying unexpected crop images and ensuring dataset integrity.