Training Machine learning model on the EuroSAT land cover classification dataset
CONSTRAINTS
- Model's inability to learn complicated features or underfitting may be caused by a lack of layers and nodes.
- The model's capacity to generalize to different datasets may be constrained by the fact that they were developed using only one dataset (EuroSAT).
- The models are trained for a set number of epochs, which might not be enough to achieve their full potential.
IMPROVEMENTS
-
To boost the performance of the models, add more layers and nodes to make them deeper and more complicated.
-
To improve feature extraction and classification from larger datasets, apply transfer learning to pre-trained models.
-
Early stopping can be used to avoid overfitting and obtain peak performance without needing to train for a set number of epochs.
-
Use methods like cross-validation to enhance the models' accuracy and prevent bias brought on by the selection of the train-test split.
-
To combine many models for improved accuracy and resilience, use ensemble learning.