Project for the university course of Vision and Cognitive Services.
We used the following datasets:
- Imagenette: https://github.com/fastai/imagenette
- Pascal: https://deepai.org/dataset/pascal-voc
- Places: https://paperswithcode.com/dataset/places205
- Birds: https://www.kaggle.com/gpiosenka/100-bird-species
- Flowers: https://www.robots.ox.ac.uk/~vgg/data/flowers/102/
Here you can find the pretrained models used in this project:
- Dahl: https://tinyclouds.org/colorize/ (Download section)
- Zhang eccv_16: https://github.com/richzhang/colorization/blob/master/colorizers/eccv16.py - related paper here
- Zhang siggraph17: https://github.com/richzhang/colorization/blob/master/colorizers/siggraph17.py - related paper here
- ChromaGAN: https://github.com/pvitoria/ChromaGAN - related paper here
- InstColorization: https://github.com/ericsujw/InstColorization - related paper here
- python 3.6 or 3.8
- virtualenv wrapper: https://virtualenvwrapper.readthedocs.io/en/latest/
-
download the pretrained model and place the
colorize.tfmodel
file in themodels
folder -
create a virtual environment:
mkvirtualenv --python=python3 dahl
-
install the requirements:
pip install -r requirements_dahl.txt
-
position yourself into the following folder:
cd src/models
-
run the model:
python3 dahl.py
-
create a virtual environment:
mkvirtualenv --python=python3 zhang
-
install the requirements:
pip install -r requirements_zhang.txt
-
position yourself into the following folder:
cd src/models
-
run the model:
python3 Eccv16andSiggraph17.py
-
download the pretrained model and place the
ChromaGAN.h5
file in themodels
folder -
create a virtual environment using python 3.6:
mkvirtualenv --python=python3 chromaGAN
-
install the requirements:
pip install -r requirements_chromaGAN.txt
-
position yourself into the following folder:
cd src/models
-
run the model:
python3 chromaGAN.py
The code to run this model is contained in the following notebook: src/models/InstColorization.ipynb
.
- AlexNet classification accuracy
- Turing test
- LPIPS: https://github.com/richzhang/PerceptualSimilarity
- PSNR and SSIM: https://cvnote.ddlee.cc/2019/09/12/psnr-ssim-python