OCR_Translator

UPDATE

2023.5.19 11:19     added spellchecker (pyenchant methods)
2023.5.21 18:56     added llama_spellcheck ([vanilla-llama](https://github.com/galatolofederico/vanilla-llama))

Introduction

This project uses EasyOCR, Page-Dewarp, Vanilla-llama and Baidu Translator for implementation, aiming to learn the encapsulation and deployment of APIs on servers

Examples

Installation

Installing on the host machine

Step1. Install OCR_Translator

git clone https://github.com/CodeRikka/OCR_Translator.git
cd OCR_Translator
pip install -r requirements.txt

Step2. Verify Page-Dewarp

page-dewarp -x 0 -y 0 pics/example6.jpg

Step3. GPU acceleration

If you wish to use GPU acceleration, please install the GPU version of torch and related packages. Check the Pytorch official website for tutorials.

pip uninstall torch torchvision torchaudio
# Modify the cuda version to your own version
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117

Step4. Download Golang related packages

go get github.com/gin-gonic/gin

Step5. Modify the filePath and outputPath in the main.go to an absolute address(optional)

Install Vanilla-llama (Optional)

See Vanilla-llama for tutorial

Run locally

Usage

usage: python export.py [-r0 ROOT_0] [-r1 ROOT_1] [-o OUTPUT_PATH]
                        [-tl THRESH_LINE] [-tb THRESH_BOX]
                        [-e EXTRA_SIZE] [-f FONT_SIZE]
                        [--uid UID] [--bkey BKEY]
                        [-g GPU{True, False}] [-rs RESIZE{True, False}]
                        [-d DEWARP{True, False}]
                        [-m MODEL{0,1,2}] # updated on 23.05.18
                        [--llama {True, False}] # updated on 23.05.21

positional arguments: [-r0 ROOT_0] [-r1 ROOT_1] [-o OUTPUT_PATH]

use Page-Dewarp: python export.py [arguments] -d True

select ML model: [-m MODEL{0,1,2}] # using ML methods
                 0: Default
                 1: KMeans
                 2: AgglomerativeClustering
                 **These methods are still being tested**
                 **See export.py for further information**

use llama_correct: [--llama True]
                   remember to start the vanilla-llama server before you use this method

Enable Server

1. Enable single threading on the local server

python single_thread.py

2. Enable multithreading on the local server

python multi_thread.py

Apply

make sure you are using the correct port

1. Send a request to the server using code

python posts/py_POST.py

2. Send a request to the server using cmd

curl -X POST https://example.com:25000/export -d root0=your/path/to/root0 -d root1=your/path/to/root1 ...