Simulated data for bangali text localisation in natural image
Description
(1) The dataset comprises paired images, consisting of styled images (i_s) and text images (i_t). The source images serve as the real background (t_b) for the target text images, and text stroke segmentation masks (mask_t) are provided to segment the text strokes in the target text images. (2) While generating synthetic data, we scale the text image in our experiment to a height of 64 pixels while maintaining the original aspect ratio. (3) The training dataset has 100,000 images in total, whereas the test dataset contains randomly 1,000 images. (4) The batch size is set to 8, and the input images are resized in size to approximately W×64 while maintaining the aspect ratio.
Files
Steps to reproduce
INSTALLATION At first, create a Conda Environment. I choose bste_dataset as the name of the environment. conda create -n bste_dataset python==3.9.0 conda activate bste_dataset pip install -r requirements.txt Note: The GitHub repository provides concise instructions on how to reproduce all the things step-by-step.