- 1.0.0
- Description
- Prepare environment.
1.1. Differently from other MapBiomas classification schemes, the Aquaculture mapping is based on DeepLearning (DL) classifier: U-Net. Thus it uses the COLAB structure, rather than purelly GEE code editor. First things first, prepare the mosaics:
users/solved/DL - Mosaic.js
- Create the water mask.
2.1. To help in the process of grabbing samples and reduce the complexity of identifying Aquaculture ponds, a water mask is generated:
users/solved/DL - Clusterization_PreDataset.js
2.2. Grabbing samples. DL classifiers demand exaustive selection of tranning samples. Guided the learning processess by manuaaly indicating areas of non-aquaculture and aquaculture:
users/solved/DL - TrainTest Geom.js
- Once trainned, run the U-Net classifier: Example:
users/solved/Mapbiomas 5 Brazil - UNet - Aquicultura.ipynb