KNN Image Classification for Land Use Dataset

This n8n workflow implements a K-Nearest Neighbors (KNN) classifier for land use image datasets, leveraging Qdrant for vector similarity search and Voyage AI for image embedding. The workflow begins by receiving an image URL via a trigger, then embeds the image using Voyage’s multimodal API. Next, it queries Qdrant to find the most similar images based on their embeddings, retrieving labeled land use classes. A majority voting mechanism determines the dominant class among the neighbors, with an intelligent loop to resolve ties by incrementally increasing the number of neighbors (limitKNN). Sticky notes provide contextual explanations for each step. This process is ideal for land classification, remote sensing, or satellite image analysis tasks, especially when analyzing large image datasets with minimal token tuning. The workflow’s flexibility allows adaptation to other image classification challenges by uploading similar datasets to Qdrant and configuring the APIs accordingly.