[2. Achieving Goal] Steps
Goal definition from [Goal Setup] Evaluation Metric:
Generate crop masks using a model given that the model achieves an F1-score of at least 0.7 on a hold out validation set of at least size 525.
- highlight low confidence pixels close to the decision boundary, e.g. pixels with scores between 0.4-0.6
- compare to other existing maps (e.g. Copernicus or GFSAD) to highlight areas where the predictions disagree and perhaps are more likely to be more difficult to classify
- randomly sample pixels or patches from the areas predicted as each class
- highlight OOD pixels within the pixels classified as each class