Induction of two different machine learning models with the purpose of classifying satellite images containing ships in two classes:
- Class 0: NO SHIP (or PARTIAL SHIP);
- Class 1: SHIP;
Possible applications of this include Ship Tracking, Ship Rescue, AShM seekers.
The two models developed are respectively:
- Single hidden-layer Neural Network;
- Support Vector Machine (SVM)
The training & evaluation of the models is perfermed on a K-Fold of the dataset, in order to get more reliable evaluation results.
After the optimization of the models it's performed an Ablation Study to verify that the preprocessing of the data is not creating performance losses.
Read the presentation we gave in here, or the full Project Relation here.
More informations about dataset, training and performance evaluation avaible in the folder relation of the repository (currently only in italian).