Download
ArtLine is a deep learning-based project focused on generating high-quality line art portraits from input images. It leverages neural network techniques built on top of the fastai library and PyTorch to transform photographic portraits into stylized line drawings. ArtLine is trained using datasets such as APDrawing and anime sketch colorization pairs to better understand facial structures and artistic line representation. An extended version integrates ControlNet, allowing users to guide the output style through textual instructions alongside the input image. ArtLine is primarily distributed as Jupyter notebooks, making it accessible for experimentation and interactive usage, especially in notebook-based environments. While the system can produce impressive results, it is sensitive to factors like lighting, background complexity, and image quality, and still struggles with elements such as shadows and fine details like hair.
Features
- Converts portrait images into stylized line art using deep learning
- Supports guided image generation using text instructions via ControlNet
- Built on fastai and PyTorch for model training and inference
- Uses curated datasets combining portrait and sketch-style images
- Available as interactive Jupyter notebooks for easy experimentation
- Capable of producing artistic outputs for creative and design workflows
Project Samples
