CNN_Keras

Implementation of a CNN in Keras.

Dataset:boats-types-recognition

First model: boat.

Use the 9 class of the dataset with 2 convolutional layers (3,3), 2 max pooling layers (2,2), 2 dropout and 2 dense layers.

Optimization function:Adadelta; Learning rate:default; Batch_size:20; Epochs:50; Steps_per_epochs:50; Validation_steps:13; Max acc:0.8428; Max val_acc:0.6618;

Second model: boat_reduced.

Use 4 class of the dataset with 3 convolutional layers (one with 5,5 and the others with 3,3), 2 max pooling layers (2,2), 2 droput and 2 dense layers.

Optimization function:Adam; Learning rate:0.001; Batch_size:16; Epochs:50; Steps_per_epochs:53; Validation_steps:11; Max acc:0.9257; Max val_acc:0.7479;

Third model: boat_reduced_2.

Use 4 class of the dataset with 3 convolutional layers (one with 5,5 and the others with 3,3), 2 max pooling layers (2,2) and 2 dense layers.

Optimization function:RMSprop; Learning rate:0.0002; Batch_size:32; Epochs:50; Steps_per_epochs:26; Validation_steps:11; Max acc:0.9670; Max val_acc:0.7596;