Merge branch 'master' of https://github.com/acmucsd/acm-ai-workshops

Author: catherinelee274

Fall20/NeuralNetworks4/layers.py

@@ -1,30 +1,33 @@
 from tensorflow.keras import Model, layers, Input
 import numpy as np
 
-temp_im_test = np.ones((1,128,128,3))
+temp_im_test = np.random.rand(1,64,64,3)
 
-input_shape_img = (128,128,3)
+input_shape_img = (64,64,3)
 
 img_input = Input(shape=input_shape_img)
 
-# convolutional layers with 64 neurons and (3,3) kernel each
-conv1_1 = layers.Conv2D(64, (3, 3), activation='relu', padding='same', name='block1_conv1')(img_input)
-conv1_2 = layers.Conv2D(64, (3, 3), activation='relu', padding='same', name='block1_conv2')(conv1_1)
+# convolutional layers with 32 neurons and (3,3) kernel each
+conv1_1 = layers.Conv2D(32, (3, 3), activation='relu', padding='same', name='block1_conv1')(img_input)
+conv1_2 = layers.Conv2D(32, (3, 3), activation='relu', padding='same', name='block1_conv2')(conv1_1)
 
 # maxpool2D layer with (2,2) filter and stride (2,2)
-max1 = layers.MaxPool2D((2, 2), strides=(2, 2), name='block1_pool')(conv1_2)
+max1 = layers.MaxPool2D((2, 2), strides=(2, 2), name='block1_max_pool')(img_input)
+
+# avgpool2D layer with (2,2) filter and stride (2,2)
+avg1 = layers.AveragePooling2D((2, 2), strides=(2, 2), name='block1_avg_pool')(img_input)
 
 # globalmax layer
-globalmax1 = layers.GlobalMaxPool2D()(max1)
+globalmax1 = layers.GlobalMaxPool2D()
 
 # flattens input to 1D array
-flatten1 = layers.Flatten()(globalmax1)
+flatten1 = layers.Flatten()(avg1)
 
 # dropout layer modifies input layer so some values become 0
 drop1 = layers.Dropout(.2)(flatten1)
 
 # standard dense layer
-dense1 = layers.Dense(32)(drop1)
+dense1 = layers.Dense(10)(drop1)
 
 # compiles model with input: img_input and output as output from dense1
 model = Model(inputs = img_input, outputs= dense1)
@@ -35,5 +38,7 @@
 
 # show result on one input
 values = model.predict(temp_im_test)
-print(values.shape)
-print(values)
+print("shape of input" + str(temp_im_test.shape))
+print("shape of output" + str(values.shape))
+print(temp_im_test[0])
+print(values[0])