mesamengistu/Habesha-Kemis-design-
Créateur:
mes
Hôte:
The objective of the project is to generate Ethiopian Habesha Kemisby using style Generative Adversarial Networks 2 (styleGAN2-ada).
## StyleGAN2 — Official TensorFlow Implementation
## Output for single image projection
## Output from random seed value
**Analyzing and Improving the Image Quality of StyleGAN**
Tero Karras, Samuli Laine, Miika Aittala, Janne Hellsten, Jaakko Lehtinen, Timo Aila
Paper: arxiv.org
Video: youtu.be
Habesha Kemis is the traditional attire o Habesha women which have a white ankle-length dress with intricate embroidery which are worn by Ethiopian women at formal events, holidays and invitations. It is made by waving shiny threads called Tilet into white fabrics that create those elegant effects. Many women also wrap a shawl called a Netela around the formal dress and the hem of the dress is quite ornated by the Tilet where fashion design is used.
In the proposed system different methods and techniques are usedto reduce the time and complexity ofdesign by implementing a system like we have developed which is called “Deep Learning based Habesha Kemisdesigning”. After reading and conceptualizing the available related systems, we have selectedstylegan2-ada to implement this project. Initially, different Habesha Kemis images are collected from many sources which are pre-processed to make them all have the same resolution and image format. After that its background is removed then converted to TensorFlow record format which are fed to the selected network structure. Generally, we used pre-trained modelssince its need less computational resources and training time. After the images are fully trained, we tasted our model by feeding input image and it successfully come up with a new Habesha Kemis design
For business inquiries, please contact researchinquiries@nvidia.com
For press and other inquiries, please contact Hector Marinez at hmarinez@nvidia.com
**★★★ NEW: StyleGAN2-ADA-PyTorch is now available; see the full list of versions here ★★★**
| Additional material |
| :--- | :-------- …