Academic project of an application that converts black and white videos or images into colored and nice ones. For more, check our Project Proposal and Video
 |
 |
|---|
A simple website containing a colorization app for video or image
-
The front-end is in HTML5, CSS3, Javascript, Ajax and Bootstrap 4.
-
Model Architectures are written using Keras.
-
The Core project is developed using Tensorflow and OpenCV.
-
Models Training is done using Colab
-
Development is OS independent.
-
We learned 3 different architectures you can find all details here.
-
By comparing the initial results of them the one which gave the best result was U-Net so we continue training on it.
-
We trained it by 66k images from old colored movies. you can find the dataset here.
-
The training took 2 days on colab to make 20 epochs with batch size 32.
- From GitHub: Clone or Download the repository or
- From Git:
- Create a new folder in the repo and name it "models"
- In this folder(models), download the model from here
- Using command line:
pip install -r requirements.txt
- Go to the project directory
- Go to Web App
- Run command line there and type:
python abbas.py
-
Include a url or upload an image then click Colorize:
-
Upload a video then click Colorize:
