Neural Style Transfer

Transform any photo into a work of art.
Apply the artistic style of famous paintings to your own images using deep neural networks. Train on your own art collection or use pretrained weights to get started instantly.

---

What It Does

Pick a content image (a photo) and a style image (a painting, artwork, etc.) — the model renders the photo in the artistic style, preserving the original composition while adopting the brushstrokes, colors, and textures of the artwork.

Highlights

• Custom training — Train on your own curated art dataset using the provided notebook
• Pretrained weights — Download pretrained weights to start stylizing immediately without training
• Comparative analysis — Built-in comparison between custom-trained and ImageNet-pretrained VGG19 weights
• Image preprocessing — Includes a resize utility notebook for preparing custom datasets (224x224 input size)

Get Started

1. Clone the repo and download the pretrained weights
2. Place the weights file in the root directory
3. Open self_trained_best.ipynb and run it

To train on your own dataset, place images in the data/ folder (one folder per style class) and run the training notebook.

Built With

Python · TensorFlow · VGG-19 · Kaggle Art Dataset

License

MIT