Similarity search: Take your search engine to the next level
summary
Explains how images become 1000‑dimensional feature vectors via a CNN (example ResNet pretrained on ImageNet) converting ~1M+ pixel RGB arrays into compact descriptors, then stores those vectors in OpenSearch (with the k-N plugin) and uses approximate k‑NN (example k=5) to return nearest images. Architecture uses SageMaker endpoints for feature extraction, OpenSearch + S3 for storage, and Lambda + API Gateway for serverless retrieval, with tunable sampling to trade accuracy for latency.
AI-written summary · about
chapters (6)
- 0:00Start
- 2:08Text Search: Analysis
- 5:03Image Search: Analysis
- 11:51Image search pipeline
- 13:33Amazon OpenSearch Service
- 18:38Implementation
related talks
22mVector Search Made Simple: Getting Started With OpenSearch for AI Applications
48mUnlocking Next-gen Semantic Search With OpenSearch 3.x
55mVector Search Made Simple: Getting Started with OpenSearch for AI Applications
30mVector Search Made Simple: Getting Started with OpenSearch for AI Applications
45mRevolutionizing Search with Power of Neural Sparse and Dense Vectors
26mOpenSearch and the Path to a Trillion-s...
43mOpenSearch and the Path to a Trillion-scale Vector Database
28mVisual Search