🎨 PICASSIFY: High-Performance Neural Style Transfer Engine

A professional-grade Computer Vision dashboard built with Python and Streamlit. This tool utilizes Neural Style Transfer (NST) via the VGG19 architecture to blend the content of one image with the artistic style of another, transforming digital photos into structured masterpieces.

---

🚀 Live Demo

Click here to try the Live App

---

✨ Features

• Neural Style Transfer: Implements the landmark Gatys et al. algorithm to decouple and recombine content and style representations.
• VGG19 Feature Extraction: Leverages a pre-trained 19-layer Convolutional Neural Network (CNN) to capture deep hierarchical features of artistic brushstrokes.
• Customizable Hyperparameters: Real-time control over style-to-content weight ratios and optimization steps for granular artistic control.
• Automated Preprocessing: Intelligent image resizing and ImageNet-standard normalization pipeline (Mean/Std-dev correction).
• Interactive UI: A dark-themed, industrial dashboard designed for high-speed local or cloud-based artistic inference.

---

🛠️ Tech Stack

• Language: Python 3.12
• Framework: Streamlit (Web UI)
• Deep Learning Engine: PyTorch / Torchvision
• Architecture: VGG19 (Deep Feature Extractor)
• Optimization: Adam Optimizer
• Image Processing: Pillow & NumPy

---

🚀 Installation & Local Setup

1. Clone the repository:

   git clone [https://github.com/ali-faraz-py/Picassify.git](https://github.com/ali-faraz-py/Picassify.git)
   cd Picassify
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Run the application:

   streamlit run app.py
   

---

📂 Project Structure

picassify/
├── app.py              # Streamlit Web UI and visualization logic
├── style_transfer.py   # Core NST engine and VGG19 feature extraction
├── utils.py            # Image processing, normalization, and tensor conversion
├── requirements.txt    # Project dependencies (PyTorch, Streamlit, etc.)
├── .gitattributes      # GitHub language statistics and configuration
├── .gitignore          # Prevents tracking of __pycache__ and local assets
└── README.md           # Project documentation

---

🧠 Model Insights

The engine utilizes a VGG19 (Visual Geometry Group) network, a powerhouse for feature representation in Computer Vision.

• The Architecture: The model treats style as a mathematical distribution of features. By calculating the Gram Matrix of specific layers, Picassify captures the texture and color patterns of the style image.

• Loss Optimization: The system simultaneously minimizes two loss functions: Content Loss (ensuring the objects remain recognizable) and Style Loss (ensuring the texture matches the artwork).

• Feature Layers: We extract content features from the deeper conv4_2 layer and style features from multiple layers (conv1_1 through conv5_1) to capture both fine detail and global artistic structure.

---

👤 Author

Syed Ali Faraz - GitHub Profile

If you found this NLP pipeline useful, please give the repository a ⭐!