OCR0058: Implementation of cGAN to generate glyph variants.

[10kalden]

Description:

• Train a conditional GAN to generate Tibetan glyphs using conditional glyphs from another font the model uses as input.
• Glyph generation with cGAN involves creating new glyphs in a specified font style based on input glyphs from a different font.
• The process uses two neural networks: a generator and a discriminator.
• The generator creates new glyphs that combine the structure of input glyphs with the stylistic features of the target font, while the discriminator evaluates these glyphs for authenticity. Through adversarial training, the generator learns to produce realistic glyphs that closely resemble the target font style.

Implementation:

Subtask:

• [x] Prepare the dataset from the publisher (test with one)
• [x] Define the model architecture
• [x] Train the model using the prepared dataset
• [ ] Tuning and make adjustment to model architecture
• [ ] Test the model

Completion Criteria To develop a model that can generate glyphs of required style

[10kalden]

• trained the model on 5k derge dataset for 30 epochs img.1 generated derge glyphs

need to do some hyperparameter tuning and modify the network architecture