Object Classification with Elasticsearch and Cloudflare

Object Classification and Search Indexing Workflow Overview This workflow automates the process of analyzing an image to identify objects within it, cropping each identified object into its own separate image file, and then indexing these new images into a searchable database. The primary goal is to build a powerful image search engine where users can find images based on the specific objects they contain, rather than relying on manual tags or filenames. Technologies Used - n8n: The core automation platform that coordinates the entire process. - Cloudflare Workers AI: Provides the AI model for object detection in images. - Cloudinary: A cloud-based service used here for hosting the cropped object images. - Elasticsearch: A powerful search and analytics engine used to store and index the data about the cropped images, making them searchable by their object label. Automation Purpose The workflow automates a complex content enrichment and management pipeline. It takes a single source image and: 1. Analyzes its content to understand what's inside. 2. Extracts individual objects as distinct assets. 3. Stores these new assets in a cloud service. 4. Indexes them in a database so they can be easily found later. This provides immense value by programmatically creating a rich, searchable library of image assets from a larger source image, saving countless hours of manual identification, cropping, and data entry. AI Integration Analysis Artificial intelligence is the central engine of this workflow, making the entire process possible. How AI is Integrated 1. Fetch Image Data: The workflow begins by retrieving an image. 2. Cloudflare AI Request: This image data is sent to a Cloudflare Workers AI endpoint, specifically using the @cf/facebook/detr-resnet-50 model. This is a pre-trained model designed for Object Detection. 3. Intelligent Analysis: The AI model processes the image and returns a structured JSON object containing a list of all the objects it identified. For each object, it provides: - A label (e.g., "car", "dog", "person"). - A score indicating the model's confidence in its identification.