pybind11-Scene-Text-Image-Transformation

Contributions

Codes borrows from

• Canjie-Luo/Scene-Text-Image-Transformer

• cxcxcxcx/imgwarp-opencv

• yati-sagade/opencv-ndarray-conversion

• edmBernard/pybind11_opencv_numpy

• cv_mat_to_numpy

A tool for scene text data augmentation. provided the tool to avoid overfitting and gain robustness of models.

Requirements

• GCC 4.8.*
• Python 3.5.*
• OpenCV 2.4.*
• PyBind11

Installation

• install OpenCV and python3 on ubuntu 16.04

• try docker Recommend✔

docker push allenzhou1996/mls-pybind11:latest

Build project:

./build.sh

python test_mls_pybind11.py

Demo

• Original Image

• Distortion

• Stretch

• Perspective

Speed

To transform an demo image with size (H:101, W:842), it takes less than 2.4ms (0.0024 sec) using a i5-8400 @2.8GHz CPU.

Citation

@inproceedings{schaefer2006image,
  title={Image deformation using moving least squares},
  author={Schaefer, Scott and McPhail, Travis and Warren, Joe},
  booktitle={ACM transactions on graphics (TOG)},
  volume={25},
  number={3},
  pages={533--540},
  year={2006},
  organization={ACM}
}