jiggum / thanos-server

Flask server for

THANOSSNAP.ORG

Frontend repo is here

Process

Instance Segmentation & Generative Inpainting

Example Cases

Before you start

Locate inpaint directory under the model_logs directory

Run

1. Run your virtualenv with python3

2. Install dependency with pip install -r requirements.txt

3. Start Flask server with S3_BUCKET_NAME='YOUR_BUCKET_NAME' python app.py

Project Structure

app.py

Contain Flask's endpoint handler. Integrating and Image save process are located.

segmentation.py

Contain instant segmentation logic with torchvision. This module generate four image.

1. Persons erased image
2. Whole persons mask
3. Half of persons mask
4. Another half of persons mask

inpaint.py

Contain inpainting logic with tensorflow. From the segmentation module's '1. Persons erased image' and '2. whole persons mask', process inpainting logic and return output image.