Official pytorch implementation of the paper: "Oil Spill Detection with A Multiscale Conditional Adversarial Network under Small Data Training". The codes will be kept updating and improving.
MCAN is an effective tool to train a model with small data.
python -m pip install -r requirements.txt
This code was tested with python 3.6
To train MCAN model on your own training images, put the training images under Input/TrainingSet, and run
python main_train.py
To test MCAN model on your own test images, put the test images under Input/TestSet, and run
python test.py --trained_model model