matlab-ocr

What problem does it solve?

This Skill helps you extract reliable text from images by guiding you through an OCR-ready pipeline in MATLAB that handles skew, polarity, preprocessing, text detection, and confidence-based validation.

Core Features & Use Cases

• End-to-end OCR pipeline: Build a complete workflow using MATLAB’s Computer Vision ocr() with a structured plan: diagnose the image, preprocess, detect text regions (e.g., CRAFT), recognize, and validate with CER/WER.
• Scene-text detection support: Use detectTextCRAFT as the default for natural scene text and support ROI-based recognition with correct LayoutAnalysis.
• Special-case routing: Handle difficult scenarios like stamped/embossed text (local contrast subtraction), text on textured backgrounds (SAM via imsegsam), inverted polarity (imcomplement), and tiny text (imresize).
• Safety/quality guardrails: Enforce critical rules such as producing an ## OCR Plan before any MATLAB calls, checking add-on availability with exist(), and saving outputs to disk before presenting extracted text.

Quick Start

Ask the AI to read text from your uploaded image and save the results by running the full MATLAB OCR pipeline using this skill.