- Dataset 1: https://data.mendeley.com/datasets/gjmr63rz2r/1
- Dataset 2: https://drive.google.com/drive/folders/1HznoBFEd6yjaLFlSmkUGARwCUzzG4whq?usp=sharing
Meg:
- Change hyperparameters of SVM-CNN hybrid model to improve metrics
- Look at implementing an SVM model
- See if you can bias the output towards being risk adverse
Charlie
- Report f1 metric for each model
- Identify what images the models are missclassifying (sunsets?)
- Test the TL models on another dataset
- Look at recall, precision, AUC metrics instead of/as well as accuracy
- Implement and compare different trasnfer learning models
- Look into which ones would be 'light-weight' for remote sensing devices