vihari / crossgrad

This repository contains implementation of CROSSGRAD (https://openreview.net/forum?id=r1Dx7fbCW) and DAN (https://arxiv.org/abs/1505.07818).

This repository contains implementation of CrossGrad (https://openreview.net/forum?id=r1Dx7fbCW) and DAN (https://arxiv.org/abs/1505.07818).

Disclaimer

The following software is shared for educational purpose only.

Dataset

We generated and used a character dataset using several hand-written fonts downloaded from Google Fonts. This dataset is referred to as GFonts dataset and is described in the CrossGrad paper further.

We make available this dataset through this repository. It can be found in the data/gfonts folder as numpy binaries.

• all_images.npy is a numpy array containing all the images in the dataset
• all_labels.npy numpy array of class labels in the same order as all_images.npy
• all_domains.npy numpy array contains the domain labels again in the same order as all_images.npy

Following is a sprite of this dataset for character: 'A'