kalapaocr 0.0.6

A boiler plate to make pip package for AI model

KALAPA OCR Package

This is a Package kalapaocr Model ONNX. The package will simply load onnx model and init an ocr engine. Give an image path and the engine will return text result.

Install Locally

After git clone, you can access the codebase and simply run the following command line:

conda create -n kalapa_env python=3.8
conda activate kalapa_env
pip install -e .

After installed, you can import kalapaocr package at anywhere when you are in kalapa_env environment

Download models

python src/kalapaocr/tool/model_downloads.py -p cached/

Basic Usage

After installing, You can view examples/sample.py to get usage of kalapaocr lib

You can run examples/sample.py file as following:

python example/sample.py -cnn cached/cnn.onnx -en cached/encoder.onnx -de cached/decoder.onnx -i 'image/test.jpg'

You can run examples/create_submission.py file as following for creating submission:

python example/create_submisson.py -cnn cached/cnn.onnx -en cached/encoder.onnx -de cached/decoder.onnx -i "your local data path/OCR/public_test" -o "your local path/your file name.csv"

Config TextRecognitor

You can initialize TextRecognitor from kalapaocr.

Let us show you:

from kalapaocr import TextRecognitor
# Using Deep Learning Models to Extract ocr results
predictor = TextRecognitor(
                        cnn_path="your cnn model path",
                        encoder_path="your encoder model path",
                        decoder_path="your decoder model path",
                    )
img = cv2.imread("your image path")
s = predictor(img)
print(s)