Extracting Structured Data from Images with OCR and LLM
summary
Builds a Deno TypeScript receipt-scanner API that uses Tesseract.js to OCR receipt images, passes the raw text to OpenAI function calling with JSON mode and a defined Deno TypeScript JSON schema (items: description, unit price, amount; subtotal and total numbers), runs a live demo via curl, returns valid JSON, and mostly extracts fields correctly though one line-item amount is wrong; validation and token optimization are explicitly omitted.
AI-written summary · about
chapters (7)
- 0:00Introduction and Speaker Background
- 0:32Understanding OCR: Basics and Importance
- 1:59Combining OCR with LLMs for Structured Data
- 2:48Demo Setup: Building the Application
- 5:20Implementing OCR with Tesseract.js
- 8:12Integrating OpenAI for Data Structuring
- 11:57Final Testing and Conclusion
related talks
01h 00mComplex Data Ingestion with Open Source AI
44mExtracting Structured Data from Legal Documents
42mExtract information accurately from documents using computer vision
36mCrash Course in Machine Learning for Natural Language Processing
42mUsing Deep Learning to Understand Documents
19mDocument Reader using Deep Learning with a Small Dataset
01h 30mUnlock Information from Tables, Images and Complex Docs
40mDeep Learning and Optical Character Recognition