EasyOCR: Optical character recognition for 80+ languages
PyTorch OCR library using CRAFT and CRNN models.
Learn more about EasyOCR
EasyOCR is a Python-based OCR library built on PyTorch that uses deep learning models for text detection and recognition. The system employs CRAFT for text detection and CRNN with LSTM layers for character recognition, supporting 80+ languages across various writing systems including Latin, Chinese, Arabic, Devanagari, and Cyrillic. Model weights are automatically downloaded on first use and cached locally, with support for both GPU and CPU inference modes. The library is commonly used for document processing, scene text recognition, and information extraction from images in production and research applications.
80+ Language Support
Single unified API handles multiple writing systems including Latin, Chinese, Arabic, Devanagari, and Cyrillic. Language models can be combined for simultaneous multi-language recognition with automatic downloading and local caching.
Flexible Input Handling
Accepts file paths, OpenCV objects, numpy arrays, and image URLs without conversion code. Returns configurable output formats—detailed results with bounding boxes and confidence scores, or simplified text-only extraction.
CPU and GPU Inference
Runs in GPU-accelerated or CPU-only modes for deployment flexibility across hardware constraints. Models load once per session and remain cached in memory for fast subsequent inference calls.
import easyocr
reader = easyocr.Reader(['en'])
result = reader.readtext('invoice.jpg')
for detection in result:
bbox, text, confidence = detection
print(f'{text} (confidence: {confidence:.2f})')Maintenance release addressing compatibility issues; release notes do not specify affected components or dependencies.
- –Review your environment for compatibility breaks, as the notes provide no detail on which libraries or Python versions are affected.
- –Test thoroughly before deploying, since the scope and nature of the fixes are not documented in the release.
Maintenance release addressing compatibility issues; release notes do not specify affected components or breaking changes.
- –Review your environment for compatibility regressions if upgrading from 1.7.0 or earlier.
- –Release notes do not detail which dependencies or platforms were fixed.
Adds native Apple Silicon support and resolves compatibility issues; no breaking changes specified.
- –Install on M1/M2 Macs without Rosetta emulation using native ARM64 wheels.
- –Release notes do not specify which compatibility fixes were applied or their impact.
Related Repositories
Discover similar tools and frameworks used by developers
Wan2.1
Diffusion transformer models for text and image-to-video generation.
sglang
High-performance inference engine for LLMs and VLMs.
unsloth
Memory-efficient Python library for accelerated LLM training.
ultralytics
PyTorch library for YOLO-based real-time computer vision.
Kimi-K2
Trillion-parameter MoE model with Muon-optimized training.