Mayank Kumar Singh

Hello @theahmedmustafa . Text annotations are required for weak supervision as it uses the number of characters in the word to break the word bounding box into equal sized segments....

Hello @yeonsikch . The ICDAR13 model is first trained on the SynthText dataset using strong supervision and later fine-tuned using weak supervision on the ICDAR 2013 dataset. As the SynthText...

Hello @malodhi . Could you check this stack overflow link? https://stackoverflow.com/questions/24791987/why-do-i-get-pickle-eoferror-ran-out-of-input-reading-an-empty-file

Hello @yeonsikch. The model file that I shared for IC13 dataset is trained on using just 1 iteration. Can you try with --iterations 1

A working example of creating the Gaussian heat map with perspective transform ---------------------------------------------------------------------------------------------------------- UPDATE 27-06-19 Added functionality to incorporate out of bound character bbox ---------------------------------------------------------------------------------------------------------- UPDATE 22-08-19 Checking if character...

> > A working example of creating the Gaussian heat map with perspective transform > > ``` > > from torch.utils import data > > import matplotlib.pyplot as plt >...

``` if np.any(bbox < 0) or np.any(bbox[:, 0] > image.shape[1]) or np.any(bbox[:, 1] > image.shape[0]): return image ``` You can add this line in the add_character function

``` def add_character(self, image, bbox): top_left = np.array([np.min(bbox[:, 0]), np.min(bbox[:, 1])]).astype(np.int32) if top_left[1] > image.shape[0] or top_left[0] > image.shape[1]: # This means there is some bug in the character bbox...

Welcome @YoungminBaek , @tinhchuquang ! @YoungminBaek isn't my code for creating word start and end points the same as yours in functionality?(Though your seems a bit cleaner!) ``` for no,...

> @mayank-git-hub > How appropriate is the value of spread? > I think 1 is the most similar to the picture in the paper. > what do you think..? >...