Welcome to Shoes Detection Project (Thai / Non-Latin Language)

Objective

detect the color of the shoes and text below it from the image with an accuracy of 99 percent by applying object detection and optical character recognition

Optical Character Recognition

Main Procedures

• synthesizing the image by utilizing synthtiger (https://github.com/clovaai/synthtiger)
• train it with deep text recognition benchmark (https://github.com/clovaai/deep-text-recognition-benchmark?tab=readme-ov-file)

Synthesizing

• follow the instructions in synthtiger GitHub page including installing fonts and corpus for your preferred language (don't forget the extract the font)
• setup and activate conda env by installing the necessary library (Pillow and raqm)
• to install raqm, you need to follow the pillow document (https://pillow.readthedocs.io/en/stable/installation/basic-installation.html), and then check the of raqm availbility by running python3 -m PIL or check.py file
• clone the fork GitHub repository of synthtiger (https://github.com/eikaramba/synthtiger/tree/pillow10), and make sure to clone it from branch pillow 10. Subsequently, use this command export PYTHONPATH=/path/to/your/forked/synthtiger:$PYTHONPATH to set it as your default Synthtiger Library in the env.
• lastly, ensure that all the files are in the correct path, and then run command synthtiger -o results -w 4 -v examples/synthtiger/template.py SynthTiger config.yaml

Training

• generating two types of data both ST and MJ.

MJ

• download preferred fonts and datasets in form of txt file
• MJ is generated by running easy_ocr_text_train/main.py

ST

• this is acquired from the synthesizing method

create data_lmdb_release by converting the folder data, containing the image and its label, to the mdb file.

• the folder structure for data_lmdb_release:

data_lmdb_release
'--- training
'       '--- MJ
'       '    '--- MJ_test
'       '             '--- data.mdb
'       '             '--- lock.mdb
'       '     '--- MJ_train
'       '             '--- data.mdb
'       '             '--- lock.mdb
'       '     '--- MJ_valid
'       '             '--- data.mdb
'       '             '--- lock.mdb
'       '--- ST
'            '--- data.mdb
'            '--- lock.mdb
'
'--- validation
         '--- data.mdb
         '--- lock.mdb

training step

requirements

• pip3 install lmdb pillow torchvision nltk natsort required
• python version 3.6 or 3.7 required

to be continued

Object Detection

to be continued

Support

If you have any question, feel free to reach out in discussion section.