EasyOCR - Optical Character Recognition

EasyOCR is a versatile OCR solution developed by JaidedAI that supports over 80 languages. It excels at detecting and recognizing text in natural scenes and real-world images.

Overview

Resources: GitHub Repository

EasyOCR provides robust text detection and recognition with extensive language support. Key features include:

• 80+ languages - Extensive multi-language support
• Scene text optimized - Excellent for real-world images (signs, labels, products)
• Two-stage pipeline - Separate detection and recognition for flexibility
• Easy to use - Simple API with good defaults
• Active community - Well-maintained with regular updates

OCR Type: Structured OCR

Structured OCR detects text regions in an image and recognizes the text within each region, returning both the text content and bounding box coordinates for each detected text block.

When to Use EasyOCR

• ✅ Scene text - Signs, billboards, product labels, street signs
• ✅ Multi-language text - Documents with multiple languages
• ✅ Natural images - Photos with text in various orientations
• ✅ Curved or rotated text - Text that isn't perfectly horizontal
• ✅ Text localization needed - When you need to know where text appears

When to Use Other OCR Models

• DocTR: Better for structured documents (invoices, forms, scanned pages)
• TrOCR: Better for single-line, pre-cropped text (serial numbers, labels)

License

Apache 2.0

Open Source License

EasyOCR is licensed under Apache 2.0, making it free for both commercial and non-commercial use without restrictions.

Learn more: Apache 2.0 License

Pre-trained Model IDs

Pre-trained EasyOCR models are available for various languages via the Roboflow API and require a Roboflow API key.

Getting a Roboflow API Key

To use EasyOCR models, you'll need a Roboflow account (free) and API key.

EasyOCR model IDs combine a recognition model (language) and a detection model using the format: easy-ocr-{language}/{detection}

Recognition Models (Languages)

Recognition models read text in specific languages.

Language	ID Chunk	Supported Languages
English	english	English
Latin	latin	English, Spanish, French, Italian, Portuguese, German, Polish, Dutch, Latin
Simplified Chinese	zh-sim	English, Simplified Chinese
Japanese	japanese	English, Japanese
Korean	korean	English, Korean
Telugu	telugu	English, Telugu
Kannada	kannada	English, Kannada
Cyrillic	cyrillic	Russian, Ukrainian, Bulgarian, Serbian, and other Cyrillic scripts

Detection Models

Detection models locate text regions in the image.

Model	ID Chunk	Description
CRAFT	craft	Character Region Awareness for Text detection
DBNet18	dbnet18	Differentiable Binarization with ResNet18 backbone
DBNet50	dbnet50	Differentiable Binarization with ResNet50 backbone

Example Model IDs

Combine any language with any detection model:

• easy-ocr-english/craft - English + CRAFT detection
• easy-ocr-latin/dbnet50 - Latin languages + DBNet50 detection (highest accuracy)
• easy-ocr-zh-sim/craft - Simplified Chinese + CRAFT detection
• easy-ocr-japanese/dbnet18 - Japanese + DBNet18 detection (balanced)

Total combinations: 8 recognition models × 3 detection models = 24 possible model configurations

Supported Backends

Backend	Extras Required
torch	torch-cpu, torch-cu118, torch-cu124, torch-cu126, torch-cu128, torch-jp6-cu126

Roboflow Platform Compatibility

Feature	Supported
Training	❌ Not available for training
Upload Weights	❌ Not supported
Serverless API (v2)	✅ Deploy via hosted API
Workflows	✅ Use in Workflows via EasyOCR block
Edge Deployment (Jetson)	✅ Deploy on NVIDIA Jetson devices
Self-Hosting	✅ Deploy with inference-models

Installation

Install with PyTorch extras:

• PyTorch: torch-cpu, torch-cu118, torch-cu124, torch-cu126, torch-cu128, torch-jp6-cu126

Usage Example

import cv2
import supervision as sv
from inference_models import AutoModel

# Load EasyOCR model for English with CRAFT detection
model = AutoModel.from_pretrained(
    "easy-ocr-english/craft",
    api_key="your_roboflow_api_key"
)

# Load image
image = cv2.imread("path/to/image.jpg")

# Run inference - returns (texts, detections)
texts, detections = model(image)

# Print detected text
for text in texts:
    print(f"Detected: {text}")

# Visualize results with supervision
box_annotator = sv.BoxAnnotator()
label_annotator = sv.LabelAnnotator()

# Annotate image with bounding boxes and text labels
annotated_image = box_annotator.annotate(scene=image.copy(), detections=detections[0])
annotated_image = label_annotator.annotate(scene=annotated_image, detections=detections[0], labels=texts)

# Save or display
cv2.imwrite("output.jpg", annotated_image)

Output Format

EasyOCR returns a tuple of (List[str], List[Detections]):

• texts: List of recognized text strings, one per detected text region
• detections: List of Detections objects with bounding boxes and metadata

This structured output allows you to know both what text was detected and where it appears in the image.