model.save_weights('/home/rev9ai/aleef/Dainsta/model/beta_v0.weights.ckpt')

# Define the PointNet architecture for regression

def build_pointnet_regression_model(num_points=2048):

    inputs = keras.Input(shape=(num_points, 3))

    x = tnet(inputs, 3)

    x = conv_bn(x, 32)

    x = conv_bn(x, 32)

    x = tnet(x, 32)

    x = conv_bn(x, 32)

    x = conv_bn(x, 64)

    x = conv_bn(x, 64) #additional

    x = conv_bn(x, 64) #additional

    x = conv_bn(x, 512)

    x = layers.GlobalMaxPooling1D()(x)

    x = dense_bn(x, 256)

    # x = layers.Dropout(0.3)(x)

    # x = dense_bn(x, 128)

    # x = layers.Dropout(0.3)(x)

    # Regression output layer

    outputs = layers.Dense(1, activation="relu")(x)

    model = keras.Model(inputs=inputs, outputs=outputs, name="pointnet_regression")

    return model

# Create the PointNet regression model

model = build_pointnet_regression_model(NUM_POINTS)

# Compile the model for regression

model.compile(

    loss="mean_absolute_error", 

    optimizer=keras.optimizers.RMSprop(learning_rate=0.1),

    metrics=["mean_absolute_error"],

)

model.load_weights('location/to/weights.ckpt')