0cve0/OpenMLKitOCR

🏛️ OpenMLkit OCR Models

This repository hosts a collection of highly optimized, lightweight, on-device OCR (Optical Character Recognition) models extracted from Google ML Kit APK components. These models are designed to run fully offline and are fully compatible with OpenMLkitOCR, a lightweight offline OCR engine for Python.

The repository includes:

• Text Detection Model: A Region Proposal Network (RPN) architecture (rpn_detector.tflite) that identifies text bounding boxes in image tiles.
• Text Recognition Models: Lightweight CRNN + CTC model pipelines for 15+ different languages and scripts.

---

🏛️ Supported Languages and Scripts

Below is the registry of the available models, vocabulary maps, and language model priors stored in this repository:

Code	Script / Language	Recognizer Model	Label Map File	Language Model / Priors
detector	Text Detection (All)	rpn_detector.tflite	—	—
en	Latin / English	line_recognizer.fb	LabelMap.pb	—
ru	Cyrillic / Russian	recognizer_cyrl.tflite	LabelMap_cyrl.pb	FST LM + Priors
zh	Chinese / Han (Hani)	recognizer_hani.tflite	recognizer_hani_label_map.pb	FST LM + Priors
ja	Japanese (Jpan)	recognizer_jpan.tflite	recognizer_jpan_label_map.pb	FST LM + Priors
ko	Korean (Kore)	recognizer_kore.tflite	recognizer_kore_label_map.pb	FST LM + Priors
ar	Arabic (Arab)	recognizer_arab_retrained.tflite	recognizer_arab_label_map.pb	FST LM + Priors
he	Hebrew (Hebr)	hebr.tflite	hebr_label_map.pb	Priors
ka	Georgian (Geor)	geor.tflite	geor_label_map.pb	Priors
bn	Bengali & Devanagari (Bede)	bede.tflite	bede_label_map.pb	Priors
gu	Gujarati (Gujr)	gocr_tflite_recognizer_gujr.tflite	gocr_tflite_recognizer_gujr_label_map.pb	Priors
kn	Kannada (Knda)	recognizer_knda.tflite	recognizer_knda_label_map.pb	FST LM + Priors
ml	Malayalam (Mlym)	recognizer_mlym.tflite	recognizer_mlym_label_map.pb	FST LM + Priors
ta	Tamil (Taml)	recognizer_taml.tflite	recognizer_taml_label_map.pb	FST LM + Priors
te	Telugu (Telu)	recognizer_telu.tflite	recognizer_telu_label_map.pb	FST LM + Priors
vi	Vietnamese / Latin	gocr_tflite_recognizer_latn_vi.tflite	gocr_tflite_recognizer_latn_vi_label_map.pb	Priors

---

📂 File Types Explained

1. *.tflite / *.fb (Neural Network weights):
   • detector/rpn_detector.tflite is a Convolutional Neural Network (CNN) that processes 256x256 tiles of the image and predicts text bounding boxes.
   • recognizer_*.tflite and line_recognizer.fb are CRNN (Convolutional Recurrent Neural Network) architectures that predict CTC logits for cropped text line images.
2. *_label_map.pb / LabelMap.pb (Vocabulary):
   • Binary Protobuf files mapping character indices to Unicode symbols for decoding CTC outputs.
3. *_lm.compact_fst.gz & *.syms (Language Models):
   • Compact Finite State Transducer (FST) language models and symbol mapping files. These are used for advanced Beam Search decoding, correcting spelling and character sequences based on word frequencies.
4. *_prior.pb / *_config.pb (Priors & Config):
   • Character prior probabilities and model configurations used to calibrate neural network outputs before applying the language model.

---

🚀 How to Use with openmlkitOCR

The Python package openmlkitOCR handles automatic downloading and caching of these models from Hugging Face if they are not present locally.

1. Installation

Install the library directly using pip:

pip install openmlkitOCR

2. Python Usage Example

import os
import cv2
from openmlkit import OpenMLKitOCR

# Configure the pipeline to pull models from this Hugging Face repository
os.environ["OPENMLKIT_MODEL_REPO"] = "0cve0/OpenMLKitOCR"

# Initialize the pipeline for a specific language (e.g., 'en' for English/Latin)
# This will automatically download and cache the detector and recognizer files.
ocr = OpenMLKitOCR(lang='en')

# Load an image
image = cv2.imread("test_image.jpg")

# Run OCR (detection & recognition)
results = ocr.run(image, score_threshold=0.35)

# Output the localized text bounding boxes and recognised characters
for item in results:
    print(f"Box: {item['box']} -> Text: {item['text']}")

---

⚖️ License and Disclaimer

• Software: The Python library OpenMLkitOCR is licensed under the Apache 2.0 License.
• Model Weights: The model weights and configurations in this repository are extracted from Google ML Kit APK components and are subject to Google's terms of service and license agreements. These models are intended for educational, research, and non-commercial local testing purposes.