(img_dim, nb_classes, model_name="Big_CNN")
| 60 | |
| 61 | |
| 62 | def Big_CNN(img_dim, nb_classes, model_name="Big_CNN"): |
| 63 | |
| 64 | x_input = Input(shape=img_dim, name="input") |
| 65 | |
| 66 | x = standard_conv_block(x_input, 64) |
| 67 | x = standard_conv_block(x, 64) |
| 68 | x = standard_conv_block(x, 64, pooling=True, dropout_rate=0.5) |
| 69 | |
| 70 | x = standard_conv_block(x, 128) |
| 71 | x = standard_conv_block(x, 128) |
| 72 | x = standard_conv_block(x, 128, pooling=True, dropout_rate=0.5) |
| 73 | |
| 74 | x = standard_conv_block(x, 256) |
| 75 | x = standard_conv_block(x, 256) |
| 76 | x = standard_conv_block(x, 256, pooling=True, dropout_rate=0.5) |
| 77 | |
| 78 | # FC part |
| 79 | x = Flatten()(x) |
| 80 | x = Dense(512, activation="relu")(x) |
| 81 | x = Dropout(0.5)(x) |
| 82 | x = Dense(512, activation="relu")(x) |
| 83 | x = Dropout(0.5)(x) |
| 84 | x = Dense(nb_classes, activation="softmax")(x) |
| 85 | |
| 86 | Big_CNN = Model(inputs=[x_input], outputs=[x]) |
| 87 | Big_CNN.name = model_name |
| 88 | |
| 89 | return Big_CNN |
| 90 | |
| 91 | |
| 92 | def load(model_name, img_dim, nb_classes): |
no test coverage detected