AI Image Search with Object Detection, CDN & Elasticsearch

This n8n workflow enables building a custom image search engine by automating the detection and indexing of objects within images. It starts by fetching an image from a provided URL, processes it through Cloudflare Workers AI using the Detr-Resnet-50 model to identify objects, their positions, labels, and confidence scores. Results with confidence below 0.9 are filtered out, and the refined data—including image URL and object associations—is indexed into an Elasticsearch instance for efficient querying and search.

Key benefits include rapid setup of scalable image search without custom ML infrastructure, leveraging cloud AI for accurate object classification, and Elasticsearch for robust full-text and similarity searches. It's ideal for e-commerce product catalogs, content management systems, media libraries, or any application needing visual content discoverability.

Use cases span digital asset management, where users search images by detected objects (e.g., 'red car' or 'laptop'); AI-enhanced galleries; or surveillance feeds. Extend by adding vector embeddings for semantic similarity, integrating with CDNs for image serving, or enriching with metadata like timestamps or user tags for advanced analytics.