The application uses machine learning to classify satellites based on their IQ samples. It begins by preprocessing the IQ samples and splitting them into training (80%) and testing (20%) subsets using stratified sampling. Transfer learning is applied using various pre-trained CNN models from the ImageNet dataset. The CNN models are modified with a dense classifier for ten classes and a softmax layer for satellite classification. The best performance is achieved with MobileNetV2, trained for 26 epochs using the RMSprop optimiser, CategoricalCrossentropy as the loss function, and CategoricalCrossentropy as the metric. Python, Keras
Transfer learning project
Zedda E
2022
Abstract
The application uses machine learning to classify satellites based on their IQ samples. It begins by preprocessing the IQ samples and splitting them into training (80%) and testing (20%) subsets using stratified sampling. Transfer learning is applied using various pre-trained CNN models from the ImageNet dataset. The CNN models are modified with a dense classifier for ten classes and a softmax layer for satellite classification. The best performance is achieved with MobileNetV2, trained for 26 epochs using the RMSprop optimiser, CategoricalCrossentropy as the loss function, and CategoricalCrossentropy as the metric. Python, KerasFile in questo prodotto:
Non ci sono file associati a questo prodotto.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
