Skip to content

Trinity College Dublin MAI project -- Prof. Anil Kokaram, Xin Shu

Notifications You must be signed in to change notification settings

Xin-Shu/AutomatedMovieRestoration

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

40 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Automated Movie Restoration

Trinity College Dublin MAI project -- Prof. Anil Kokaram, Research Xin Shu

Thesis Reference

This repository contains the source codes for the MAI research project specified below. The workflow contains the Aritficial degrade film dataset preparation, CNN model training, and prediction post-processing.

@techreport{AutomateLineScratchDetection,
    author = {Shu, Xin},
    title = {Automated Film Scratch Detection},
    institution = {Trinity College Dublin},
    year = {2022},
    month = {April},
}

Proposed CNN architecture

With this architecture, 3,428,642 trainable parameters and 1,408 non-trainable parameters are involved. CNN architecture

Obtained Resutls

Detection example from Cinecitta footage Detection example from the Knight Detection example from the Sit Down Detection example from the Carrier Detection example from a unknow RGB sequence

Environment Requirements

The code has been tested with Python 3.6 and TensorFlow 1.15.5 on AMD core device.

conda create --name afsd python=3.6
conda activate afsd
conda install numpy, opencv-python
pip install tensorflow-directml

Releases

No releases published

Packages

No packages published

Languages