Case study
Fast OCR ONNX Inference Server
Containerized OCR API that stages line segmentation, word segmentation, and CRNN text recognition behind a FastAPI endpoint.
Overview
I built a containerized OCR inference service that turns line segmentation, word segmentation and CRNN recognition into a FastAPI image-upload workflow. It returns recognized text with line and word boxes as structured JSON. ONNX Runtime provides a portable CPU-serving path, with Docker packaging and explicit response contracts connecting the model pipeline to downstream applications.
Project figures
What It Covers
- Line segmentation, word segmentation and CRNN text recognition
- Image-upload API returning recognized text and geometry
- Portable CPU inference with ONNX Runtime
- Docker packaging and explicit downstream response contracts
Stack And Topics
- Python
- FastAPI
- ONNX Runtime
- CRNN
- OCR
- Docker
- Cloud Run