Python Implementation of Generative Adversarial Networks for Extrapolation of Corrosion in Automobile Images

This repository is provided as a tutorial for the rust level and texture maps prediction and application to cars' fender.

The repository consists of the following four folders:

ds:

• train contains the training set used for the Rust Level Network (RLN) and Rust Texture Network (RTN) training
• test contains the images that were lately acquired for a comparison with the models' predicted maps

target-cars:

• img contains car images.
• mask contains the masks for each car image.
• annotation holds three points along the fender of each car (the start of the fender, its end, and an arbitrary point in between) and the estimated radius in pixels.

samples:

• maps contains examples of maps generated by our RLN and RTN models.
• cars contains the results of the generated maps applied to the target cars.

src:

• utils is a set of functions used for the rust map application.
• main is a back-to-back pipeline for our application, from the RLN and RTN predictions until the rust maps application in a car.
• model contains our RLN and RTN generator and discriminator architectures.
• data contains a function that wraps our dataset

weights:

• this folder contain the trained weights of our RLN and RTN generator and discriminator models

The file src/main.py is the executable file of this project.