MecatronicaUSB/uw-color-correction-ml

Underwater color correction using GANs and CNNs

Underwater Color Correction using ML

This README is under construction, but the project consists of using GANs to generate a synthetic dataset to train a CNN to correct the color of underwater images.

Python environment

If it is your first time running the project, go from step 1 to 5. If you already installed the requeriments and want to run it, just activate the environment with the step 3.

1. Install python3-venv:

sudo apt-get install python3-venv -y

2. In the root path of this repo, run:

python3 -m venv v-env

3. Activate the environment:

source v-env/bin/activate

4. Run this before requirements so it doesn't fail:

pip install wheel

5. Install the project requirements:

pip install -r requirements.txt

Setting the parameters file

1. At cnn/ and gan/ change the parameters.json.example filename to parameters.json. This is done so your local changes in the paths and hyperparameters are not uploaded to GitHub.

Running the GAN

1. Make sure your environment is activated:

source v-env/bin/activate

2. To train the GAN, run:

cd gan/
python main.py

Authors and Contributors

• Gabriel Noya - gnoya
• Jose Cappelletto - cappelletto