How tankalanagaraju/historical-photo-colorization-using-ai works
This is a Streamlit web app for turning grayscale photos into downloadable colorized images. Users upload an image or choose a sample; the app checks that it is grayscale, runs Caffe-based colorization, displays the result, and offers a PNG download. The inference stages are implemented together in colorizer() in app.py. The tree does not include the Input_images directory referenced by the sample-image selector, so that path is not represented as a repository component.
Main entry points: app.py Colorizer · models_colorization_deploy_v2.prototxt Caffe config · colorization_release_v2.caffemodel CNN weights · pts_in_hull.npy Color points
Components 10
User interaction
- Image interface
app.py - Image input
app.py - Grayscale check
app.py - PNG download
app.py
Colorization pipeline
- Colorizer
app.py - OpenCV DNN
Model assets
- Caffe config
models/models_colorization_deploy_v2.prototxt - CNN weights
models/colorization_release_v2.caffemodel - Color points
models/pts_in_hull.npy
Other components
- User
Connections 12
- User → Image interface: uses
- Image interface → Image input: accepts image
- Image input → Grayscale check: provides image
- Image interface → Grayscale check: checks image
- Grayscale check → Colorizer: colorizes grayscale
- Colorizer → OpenCV DNN: runs inference
- Colorizer → Caffe config: loads config
- Colorizer → CNN weights: loads weights
- Colorizer → Color points: loads points
- Image interface → PNG download: offers result
- Colorizer → PNG download: returns image
- User → PNG download: downloads PNG