Deep Learning with PyTorch – Image Segmentation

A lightweight SaaS web application using Streamlit allows users to upload an image, run it through the trained PyTorch model, and download the resulting images.

Tool Options Available:

• Show Mask Only: Displays the raw segmentation mask.
• Transparent BG: Removes the background.
• Portrait Mode: Applies a Gaussian blur to the background.
• B&W BG: Converts the background to black and white while keeping the subject in color.

---

🛠️ Training Process

The deep learning model is based on the U-Net architecture with an EfficientNet-b0 encoder. Using PyTorch, it was trained to map input images to binary masks, effectively learning to identify and separate the main subject from the background.

TRAINING follows the Coursera course: Deep Learning with PyTorch – Image Segmentation - (Verification ID: TQXF0B23K30A).

---

How to Run Locally

1. Activate the virtual environment:

   source venv/bin/activate

2. Install dependencies:

   pip install -r requirements.txt

3. Start the Streamlit application:

   streamlit run app.py