This project was made for HackTech 2021 by Elise Liu and Archie Shahidullah. It uses a pix2pix conditional GAN
neural network to forecast deforestation in the Amazon rainforest. The code to train this model
is in pix2pix_train.ipynb. We used a GPU on Google Colaboratory to train the model. The following command runs the flask app, where you can select a region from a map, using the Mapbox API, and forecast 5 years into the future of what deforestation may look like.
python main.py
Note we do not include the folder training_checkpoints which has the
650 Mb model weights file or the Mapbox token.txt file.
Here is a sample image to predict on.
This is the forecast returned by the model.
See our Devpost project for more information.