OCR
Optical character recognition is the automated conversion of text depicted in images or scanned documents into machine-readable characters.
Digital Twins in Action
OCR is the automated recognition and conversion of text in images or scanned documents into machine-readable characters.
- How it is used
- Recovers readings from archived bills and meter photographs for a digital twin’s historical baseline and more frequent updates.
- How deep it goes
- Works through bill parsing and photograph preprocessing, but does not explain Tesseract’s recognition internals.
- Examples
- Uses a scanned gas bill, a water-meter photograph, and a battery-powered meter camera.
- What you can do after
- Apply it via Tesseract to extract readings from preprocessed meter photos; operate an edge camera with field calibration and verification images.
Machine Learning Platform Engineering
Optical character recognition is the automated extraction of written or printed text from images.
- How it is used
- Frames OCR as a production ML use case for extracting ID numbers from identity-card images.
- How deep it goes
- Traces data collection, annotation, dataset splitting, detection, fine-tuning, and deployment without detailing recognition internals.
- Examples
- Uses identity-card images, ID-number bounding boxes, handwritten transcriptions, and country-varied card datasets.
- What you can do after
- Build an OCR service by preparing labeled images, training a detector, fine-tuning it, and deploying the result.
Read how Machine Learning Platform Engineering explains OCR →
Related concepts
- corpusproduces · A corpus is an organized collection of texts or other language data assembled for analysis, retrieval, training, or evaluation.
- feature engineeringhas part · Feature engineering is the creation, transformation, or selection of input features from data to provide useful representations for analysis or modeling.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.