In this group project, you are going to build a 3D Conv model that will be able to predict the 5 gestures correctly. Please import the following libraries to get started. Once you have completed the code you can download the notebook for making a submission.
import numpy as np
import os
from imageio import imread
from skimage.transform import resize
import datetime
import osWe set the random seed so that the results don't vary drastically.
np.random.seed(30)
import random as rn
rn.seed(30)
from tensorflow import keras
import tensorflow as tf
tf.random.set_seed(30)In this block, you read the folder names for training and validation. You also set the batch_size here. Note that you set the batch size in such a way that you are able to use the GPU in full capacity. You keep increasing the batch size until the machine throws an error.
def plot(history):
fig, axes = plt.subplots(nrows=1, ncols=2, figsize=(15,4))
axes[0].plot(history.history['loss'])
axes[0].plot(history.history['val_loss'])
axes[0].legend(['loss','val_loss'])
axes[1].plot(history.history['categorical_accuracy'])
axes[1].plot(history.history['val_categorical_accuracy'])
axes[1].legend(['categorical_accuracy','val_accuracy'])data path: /home/datasets/Project_data
train_doc = np.random.permutation(open('/home/datasets/Project_data/train.csv').readlines())
val_doc = np.random.permutation(open('/home/datasets/Project_data/val.csv').readlines())
# Our Hyperparameters
batch_size = 32 #experiment with the batch size
image_height=120
image_width=120
channels=3
num_classes=5
total_frames=30
frame_samples=16
num_epochs = 62This is one of the most important part of the code. The overall structure of the generator has been given. In the generator, you are going to preprocess the images as you have images of 2 different dimensions as well as create a batch of video frames. You have to experiment with img_idx, y,z and normalization such that you get high accuracy.
def generator(source_path, folder_list, batch_size):
print( 'Source path = ', source_path, '; batch size =', batch_size)
img_idx = np.round(np.linspace(0,total_frames-1,frame_samples)).astype(int)#create a list of image numbers you want to use for a particular video
while True:
t = np.random.permutation(folder_list)
num_batches = len(t)//batch_size# calculate the number of batches
for batch in range(num_batches): # we iterate over the number of batches
remaining_seq=0 #To track for remainning sequence
batch_data = np.zeros((batch_size,len(img_idx),image_height,image_width,3)) # len(img_idx) is the number of images you use for each video, (image_height,image_width) is the final size of the input images and 3 is the number of channels RGB
batch_labels = np.zeros((batch_size,5)) # batch_labels is the one hot representation of the output
for folder in range(batch_size): # iterate over the batch_size
imgs = os.listdir(source_path+'/'+ t[folder + (batch*batch_size)].split(';')[0]) # read all the images in the folder
for idx,item in enumerate(img_idx): # Iterate iver the frames/images of a folder to read them in
image = imread(source_path+'/'+ t[folder + (batch*batch_size)].strip().split(';')[0]+'/'+imgs[item]).astype(np.float32)
if image.shape[0] != image.shape[1]:
image=image[0:120,10:150]
else:
image=image[30:330,20:340]
image_resized=resize(image,(image_height,image_width,3))
#crop the images and resize them. Note that the images are of 2 different shape
#and the conv3D will throw error if the inputs in a batch have different shapes
batch_data[folder,idx,:,:,0] = (image_resized[:,:,0])/255 #normalise and feed in the image
batch_data[folder,idx,:,:,1] = (image_resized[:,:,1])/255 #normalise and feed in the image
batch_data[folder,idx,:,:,2] = (image_resized[:,:,2])/255 #normalise and feed in the image
batch_labels[folder, int(t[folder + (batch*batch_size)].strip().split(';')[2])] = 1
yield batch_data, batch_labels #you yield the batch_data and the batch_labels, remember what does yield do
# write the code for the remaining data points which are left after full batches
remaining_seq=len(t)%batch_size #To track for remainning sequence
if (remaining_seq != 0):
seq_len = remaining_seq #To track for remainning sequence
batch_data = np.zeros((seq_len,len(img_idx),image_height,image_width,3)) # len(img_idx) is the number of images you use for each video, (image_height,image_width) is the final size of the input images and 3 is the number of channels RGB
batch_labels = np.zeros((seq_len,5)) # batch_labels is the one hot representation of the output
for folder in range(seq_len): # iterate over the batch_size
imgs = os.listdir(source_path+'/'+ t[folder + (batch*batch_size)].split(';')[0]) # read all the images in the folder
for idx,item in enumerate(img_idx): # Iterate iver the frames/images of a folder to read them in
image = imread(source_path+'/'+ t[folder + (batch*batch_size)].strip().split(';')[0]+'/'+imgs[item]).astype(np.float32)
if image.shape[0] != image.shape[1]:
image=image[0:120,10:150]
else:
image=image[30:340,20:340]
image_resized=resize(image,(image_height,image_width,3))
#crop the images and resize them. Note that the images are of 2 different shape
#and the conv3D will throw error if the inputs in a batch have different shapes
batch_data[folder,idx,:,:,0] = (image_resized[:,:,0])/255 #normalise and feed in the image
batch_data[folder,idx,:,:,1] = (image_resized[:,:,1])/255 #normalise and feed in the image
batch_data[folder,idx,:,:,2] = (image_resized[:,:,2])/255 #normalise and feed in the image
batch_labels[folder, int(t[folder + (batch*batch_size)].strip().split(';')[2])] = 1
yield batch_data, batch_labels #you yield the batch_data and the batch_labels, remember what does yield doNote here that a video is represented above in the generator as (number of images, height, width, number of channels). Take this into consideration while creating the model architecture.
curr_dt_time = datetime.datetime.now()
train_path = '/home/datasets/Project_data/train'
val_path = '/home/datasets/Project_data/val'
num_train_sequences = len(train_doc)
print('# training sequences =', num_train_sequences)
num_val_sequences = len(val_doc)
print('# validation sequences =', num_val_sequences)
print ('# epochs =', num_epochs)# training sequences = 663
# validation sequences = 100
# epochs = 62
Here you make the model using different functionalities that Keras provides. Remember to use Conv3D and MaxPooling3D and not Conv2D and Maxpooling2D for a 3D convolution model. You would want to use TimeDistributed while building a Conv2D + RNN model. Also remember that the last layer is the softmax. Design the network in such a way that the model is able to give good accuracy on the least number of parameters so that it can fit in the memory of the webcam.
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, GRU, Dropout, Flatten, BatchNormalization, Activation, Conv3D, MaxPooling3D
from tensorflow.keras.callbacks import ModelCheckpoint, ReduceLROnPlateau
from tensorflow.keras import optimizers
#write your model here
model = Sequential()
model.add(Conv3D(16, (3,3,3), padding='same',input_shape=(frame_samples,image_height,image_width,channels)))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(MaxPooling3D(pool_size=(2, 2, 2)))
#model.add(Dropout(0.4))
model.add(Conv3D(32, (3,3,3), padding='same'))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(MaxPooling3D(pool_size=(2, 2, 2)))
#model.add(Dropout(0.4))
model.add(Conv3D(64, (3,3,3), padding='same'))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(MaxPooling3D(pool_size=(2, 2, 2)))
model.add(Dropout(0.4))
#model.add(Conv3D(128, (3,3,3), padding='same'))
#model.add(Activation('relu'))
#model.add(BatchNormalization())
#model.add(MaxPooling3D(pool_size=(2, 2, 2)))
model.add(Flatten())
model.add(Dense(256,activation='relu'))
model.add(BatchNormalization())
model.add(Dropout(0.5))
#model.add(Dense(128,activation='relu'))
#model.add(BatchNormalization())
#model.add(Dropout(0.5))
model.add(Dense(num_classes,activation='softmax'))Now that you have written the model, the next step is to compile the model. When you print the summary of the model, you'll see the total number of parameters you have to train.
optimiser = optimizers.Adam(lr=0.009)#write your optimizer
model.compile(optimizer=optimiser, loss='categorical_crossentropy', metrics=['categorical_accuracy'])
print (model.summary())Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv3d (Conv3D) (None, 16, 120, 120, 16) 1312
_________________________________________________________________
batch_normalization (BatchNo (None, 16, 120, 120, 16) 64
_________________________________________________________________
activation (Activation) (None, 16, 120, 120, 16) 0
_________________________________________________________________
max_pooling3d (MaxPooling3D) (None, 8, 60, 60, 16) 0
_________________________________________________________________
conv3d_1 (Conv3D) (None, 8, 60, 60, 32) 13856
_________________________________________________________________
batch_normalization_1 (Batch (None, 8, 60, 60, 32) 128
_________________________________________________________________
activation_1 (Activation) (None, 8, 60, 60, 32) 0
_________________________________________________________________
max_pooling3d_1 (MaxPooling3 (None, 4, 30, 30, 32) 0
_________________________________________________________________
conv3d_2 (Conv3D) (None, 4, 30, 30, 64) 55360
_________________________________________________________________
batch_normalization_2 (Batch (None, 4, 30, 30, 64) 256
_________________________________________________________________
activation_2 (Activation) (None, 4, 30, 30, 64) 0
_________________________________________________________________
max_pooling3d_2 (MaxPooling3 (None, 2, 15, 15, 64) 0
_________________________________________________________________
dropout (Dropout) (None, 2, 15, 15, 64) 0
_________________________________________________________________
flatten (Flatten) (None, 28800) 0
_________________________________________________________________
dense (Dense) (None, 256) 7373056
_________________________________________________________________
batch_normalization_3 (Batch (None, 256) 1024
_________________________________________________________________
dropout_1 (Dropout) (None, 256) 0
_________________________________________________________________
dense_1 (Dense) (None, 5) 1285
=================================================================
Total params: 7,446,341
Trainable params: 7,445,605
Non-trainable params: 736
_________________________________________________________________
None
Let us create the train_generator and the val_generator which will be used in .fit_generator.
train_generator = generator(train_path, train_doc, batch_size)
val_generator = generator(val_path, val_doc, batch_size)model_name = 'model' + '_' + str(curr_dt_time).replace(' ','').replace(':','_') + '/'
if not os.path.exists(model_name):
os.mkdir(model_name)
filepath = model_name + 'model-{epoch:05d}-{loss:.5f}-{categorical_accuracy:.5f}-{val_loss:.5f}-{val_categorical_accuracy:.5f}.h5'
checkpoint = ModelCheckpoint(filepath, monitor='val_loss', verbose=1, save_best_only=False, save_weights_only=False, mode='auto',period = 1)
LR = ReduceLROnPlateau(monitor='val_loss', factor=0.01, patience=5, cooldown=4, verbose=1,mode='auto',epsilon=0.0001)# write the REducelronplateau code here
callbacks_list = [checkpoint, LR]WARNING:tensorflow:`period` argument is deprecated. Please use `save_freq` to specify the frequency in number of batches seen.
WARNING:tensorflow:`epsilon` argument is deprecated and will be removed, use `min_delta` instead.
The steps_per_epoch and validation_steps are used by fit method to decide the number of next() calls it need to make.
if (num_train_sequences%batch_size) == 0:
steps_per_epoch = int(num_train_sequences/batch_size)
else:
steps_per_epoch = (num_train_sequences//batch_size) + 1
if (num_val_sequences%batch_size) == 0:
validation_steps = int(num_val_sequences/batch_size)
else:
validation_steps = (num_val_sequences//batch_size) + 1Let us now fit the model. This will start training the model and with the help of the checkpoints, you'll be able to save the model at the end of each epoch.
history = model.fit(train_generator, steps_per_epoch=steps_per_epoch, epochs=num_epochs, verbose=1,
callbacks=callbacks_list, validation_data=val_generator,
validation_steps=validation_steps, class_weight=None, workers=1, initial_epoch=0)Source path = /home/datasets/Project_data/train ; batch size = 32
Epoch 1/62
21/21 [==============================] - ETA: 0s - loss: 1.5595 - categorical_accuracy: 0.4208Source path = /home/datasets/Project_data/val ; batch size = 32
Epoch 00001: saving model to model_2021-06-2711_07_38.791020/model-00001-1.55945-0.42081-134.33029-0.21000.h5
21/21 [==============================] - 70s 3s/step - loss: 1.5595 - categorical_accuracy: 0.4208 - val_loss: 134.3303 - val_categorical_accuracy: 0.2100
Epoch 2/62
21/21 [==============================] - ETA: 0s - loss: 0.9805 - categorical_accuracy: 0.5943
Epoch 00002: saving model to model_2021-06-2711_07_38.791020/model-00002-0.98051-0.59427-45.58620-0.23000.h5
21/21 [==============================] - 73s 3s/step - loss: 0.9805 - categorical_accuracy: 0.5943 - val_loss: 45.5862 - val_categorical_accuracy: 0.2300
Epoch 3/62
21/21 [==============================] - ETA: 0s - loss: 0.7353 - categorical_accuracy: 0.7210
Epoch 00003: saving model to model_2021-06-2711_07_38.791020/model-00003-0.73531-0.72097-19.00770-0.27000.h5
21/21 [==============================] - 72s 3s/step - loss: 0.7353 - categorical_accuracy: 0.7210 - val_loss: 19.0077 - val_categorical_accuracy: 0.2700
Epoch 4/62
21/21 [==============================] - ETA: 0s - loss: 0.5677 - categorical_accuracy: 0.7662
Epoch 00004: saving model to model_2021-06-2711_07_38.791020/model-00004-0.56769-0.76621-15.87168-0.25000.h5
21/21 [==============================] - 70s 3s/step - loss: 0.5677 - categorical_accuracy: 0.7662 - val_loss: 15.8717 - val_categorical_accuracy: 0.2500
Epoch 5/62
21/21 [==============================] - ETA: 0s - loss: 0.5216 - categorical_accuracy: 0.8145
Epoch 00005: saving model to model_2021-06-2711_07_38.791020/model-00005-0.52156-0.81448-2.56329-0.43000.h5
21/21 [==============================] - 72s 3s/step - loss: 0.5216 - categorical_accuracy: 0.8145 - val_loss: 2.5633 - val_categorical_accuracy: 0.4300
Epoch 6/62
21/21 [==============================] - ETA: 0s - loss: 0.3695 - categorical_accuracy: 0.8567
Epoch 00006: saving model to model_2021-06-2711_07_38.791020/model-00006-0.36950-0.85671-2.18063-0.47000.h5
21/21 [==============================] - 74s 4s/step - loss: 0.3695 - categorical_accuracy: 0.8567 - val_loss: 2.1806 - val_categorical_accuracy: 0.4700
Epoch 7/62
21/21 [==============================] - ETA: 0s - loss: 0.2754 - categorical_accuracy: 0.8989
Epoch 00007: saving model to model_2021-06-2711_07_38.791020/model-00007-0.27544-0.89894-0.70012-0.72000.h5
21/21 [==============================] - 74s 4s/step - loss: 0.2754 - categorical_accuracy: 0.8989 - val_loss: 0.7001 - val_categorical_accuracy: 0.7200
Epoch 8/62
21/21 [==============================] - ETA: 0s - loss: 0.2390 - categorical_accuracy: 0.9170
Epoch 00008: saving model to model_2021-06-2711_07_38.791020/model-00008-0.23901-0.91704-2.42922-0.50000.h5
21/21 [==============================] - 69s 3s/step - loss: 0.2390 - categorical_accuracy: 0.9170 - val_loss: 2.4292 - val_categorical_accuracy: 0.5000
Epoch 9/62
21/21 [==============================] - ETA: 0s - loss: 0.2223 - categorical_accuracy: 0.9216
Epoch 00009: saving model to model_2021-06-2711_07_38.791020/model-00009-0.22235-0.92157-1.47980-0.65000.h5
21/21 [==============================] - 71s 3s/step - loss: 0.2223 - categorical_accuracy: 0.9216 - val_loss: 1.4798 - val_categorical_accuracy: 0.6500
Epoch 10/62
21/21 [==============================] - ETA: 0s - loss: 0.1823 - categorical_accuracy: 0.9442
Epoch 00010: saving model to model_2021-06-2711_07_38.791020/model-00010-0.18233-0.94419-2.57295-0.59000.h5
21/21 [==============================] - 71s 3s/step - loss: 0.1823 - categorical_accuracy: 0.9442 - val_loss: 2.5730 - val_categorical_accuracy: 0.5900
Epoch 11/62
21/21 [==============================] - ETA: 0s - loss: 0.1434 - categorical_accuracy: 0.9563
Epoch 00011: saving model to model_2021-06-2711_07_38.791020/model-00011-0.14339-0.95626-1.07276-0.72000.h5
21/21 [==============================] - 70s 3s/step - loss: 0.1434 - categorical_accuracy: 0.9563 - val_loss: 1.0728 - val_categorical_accuracy: 0.7200
Epoch 12/62
21/21 [==============================] - ETA: 0s - loss: 0.1174 - categorical_accuracy: 0.9593
Epoch 00012: saving model to model_2021-06-2711_07_38.791020/model-00012-0.11743-0.95928-1.54299-0.65000.h5
Epoch 00012: ReduceLROnPlateau reducing learning rate to 8.999999612569808e-05.
21/21 [==============================] - 68s 3s/step - loss: 0.1174 - categorical_accuracy: 0.9593 - val_loss: 1.5430 - val_categorical_accuracy: 0.6500
Epoch 13/62
21/21 [==============================] - ETA: 0s - loss: 0.0853 - categorical_accuracy: 0.9713
Epoch 00013: saving model to model_2021-06-2711_07_38.791020/model-00013-0.08530-0.97134-1.22011-0.72000.h5
21/21 [==============================] - 69s 3s/step - loss: 0.0853 - categorical_accuracy: 0.9713 - val_loss: 1.2201 - val_categorical_accuracy: 0.7200
Epoch 14/62
21/21 [==============================] - ETA: 0s - loss: 0.0763 - categorical_accuracy: 0.9759
Epoch 00014: saving model to model_2021-06-2711_07_38.791020/model-00014-0.07628-0.97587-0.76473-0.82000.h5
21/21 [==============================] - 71s 3s/step - loss: 0.0763 - categorical_accuracy: 0.9759 - val_loss: 0.7647 - val_categorical_accuracy: 0.8200
Epoch 15/62
21/21 [==============================] - ETA: 0s - loss: 0.0697 - categorical_accuracy: 0.9849
Epoch 00015: saving model to model_2021-06-2711_07_38.791020/model-00015-0.06975-0.98492-0.76628-0.78000.h5
21/21 [==============================] - 72s 3s/step - loss: 0.0697 - categorical_accuracy: 0.9849 - val_loss: 0.7663 - val_categorical_accuracy: 0.7800
Epoch 16/62
21/21 [==============================] - ETA: 0s - loss: 0.0726 - categorical_accuracy: 0.9759
Epoch 00016: saving model to model_2021-06-2711_07_38.791020/model-00016-0.07258-0.97587-0.55069-0.84000.h5
21/21 [==============================] - 67s 3s/step - loss: 0.0726 - categorical_accuracy: 0.9759 - val_loss: 0.5507 - val_categorical_accuracy: 0.8400
Epoch 17/62
21/21 [==============================] - ETA: 0s - loss: 0.1027 - categorical_accuracy: 0.9623
Epoch 00017: saving model to model_2021-06-2711_07_38.791020/model-00017-0.10273-0.96229-0.41232-0.84000.h5
21/21 [==============================] - 69s 3s/step - loss: 0.1027 - categorical_accuracy: 0.9623 - val_loss: 0.4123 - val_categorical_accuracy: 0.8400
Epoch 18/62
21/21 [==============================] - ETA: 0s - loss: 0.0499 - categorical_accuracy: 0.9910
Epoch 00018: saving model to model_2021-06-2711_07_38.791020/model-00018-0.04987-0.99095-0.49211-0.85000.h5
21/21 [==============================] - 70s 3s/step - loss: 0.0499 - categorical_accuracy: 0.9910 - val_loss: 0.4921 - val_categorical_accuracy: 0.8500
Epoch 19/62
21/21 [==============================] - ETA: 0s - loss: 0.0590 - categorical_accuracy: 0.9849
Epoch 00019: saving model to model_2021-06-2711_07_38.791020/model-00019-0.05899-0.98492-0.38509-0.87000.h5
21/21 [==============================] - 71s 3s/step - loss: 0.0590 - categorical_accuracy: 0.9849 - val_loss: 0.3851 - val_categorical_accuracy: 0.8700
Epoch 20/62
21/21 [==============================] - ETA: 0s - loss: 0.0410 - categorical_accuracy: 0.9925
Epoch 00020: saving model to model_2021-06-2711_07_38.791020/model-00020-0.04097-0.99246-0.34931-0.89000.h5
21/21 [==============================] - 67s 3s/step - loss: 0.0410 - categorical_accuracy: 0.9925 - val_loss: 0.3493 - val_categorical_accuracy: 0.8900
Epoch 21/62
21/21 [==============================] - ETA: 0s - loss: 0.0573 - categorical_accuracy: 0.9834
Epoch 00021: saving model to model_2021-06-2711_07_38.791020/model-00021-0.05735-0.98341-0.36788-0.88000.h5
21/21 [==============================] - 71s 3s/step - loss: 0.0573 - categorical_accuracy: 0.9834 - val_loss: 0.3679 - val_categorical_accuracy: 0.8800
Epoch 22/62
21/21 [==============================] - ETA: 0s - loss: 0.0687 - categorical_accuracy: 0.9774
Epoch 00022: saving model to model_2021-06-2711_07_38.791020/model-00022-0.06868-0.97738-0.43010-0.84000.h5
21/21 [==============================] - 70s 3s/step - loss: 0.0687 - categorical_accuracy: 0.9774 - val_loss: 0.4301 - val_categorical_accuracy: 0.8400
Epoch 23/62
21/21 [==============================] - ETA: 0s - loss: 0.0495 - categorical_accuracy: 0.9910
Epoch 00023: saving model to model_2021-06-2711_07_38.791020/model-00023-0.04949-0.99095-0.34236-0.88000.h5
21/21 [==============================] - 70s 3s/step - loss: 0.0495 - categorical_accuracy: 0.9910 - val_loss: 0.3424 - val_categorical_accuracy: 0.8800
Epoch 24/62
21/21 [==============================] - ETA: 0s - loss: 0.0709 - categorical_accuracy: 0.9744
Epoch 00024: saving model to model_2021-06-2711_07_38.791020/model-00024-0.07095-0.97436-0.35519-0.88000.h5
21/21 [==============================] - 67s 3s/step - loss: 0.0709 - categorical_accuracy: 0.9744 - val_loss: 0.3552 - val_categorical_accuracy: 0.8800
Epoch 25/62
21/21 [==============================] - ETA: 0s - loss: 0.0525 - categorical_accuracy: 0.9849
Epoch 00025: saving model to model_2021-06-2711_07_38.791020/model-00025-0.05250-0.98492-0.34796-0.89000.h5
21/21 [==============================] - 69s 3s/step - loss: 0.0525 - categorical_accuracy: 0.9849 - val_loss: 0.3480 - val_categorical_accuracy: 0.8900
Epoch 26/62
21/21 [==============================] - ETA: 0s - loss: 0.0545 - categorical_accuracy: 0.9819
Epoch 00026: saving model to model_2021-06-2711_07_38.791020/model-00026-0.05448-0.98190-0.35833-0.89000.h5
21/21 [==============================] - 70s 3s/step - loss: 0.0545 - categorical_accuracy: 0.9819 - val_loss: 0.3583 - val_categorical_accuracy: 0.8900
Epoch 27/62
21/21 [==============================] - ETA: 0s - loss: 0.0504 - categorical_accuracy: 0.9879
Epoch 00027: saving model to model_2021-06-2711_07_38.791020/model-00027-0.05037-0.98793-0.39564-0.88000.h5
21/21 [==============================] - 70s 3s/step - loss: 0.0504 - categorical_accuracy: 0.9879 - val_loss: 0.3956 - val_categorical_accuracy: 0.8800
Epoch 28/62
21/21 [==============================] - ETA: 0s - loss: 0.0692 - categorical_accuracy: 0.9698
Epoch 00028: saving model to model_2021-06-2711_07_38.791020/model-00028-0.06919-0.96983-0.28273-0.91000.h5
21/21 [==============================] - 67s 3s/step - loss: 0.0692 - categorical_accuracy: 0.9698 - val_loss: 0.2827 - val_categorical_accuracy: 0.9100
Epoch 29/62
21/21 [==============================] - ETA: 0s - loss: 0.0509 - categorical_accuracy: 0.9879
Epoch 00029: saving model to model_2021-06-2711_07_38.791020/model-00029-0.05087-0.98793-0.28350-0.91000.h5
21/21 [==============================] - 70s 3s/step - loss: 0.0509 - categorical_accuracy: 0.9879 - val_loss: 0.2835 - val_categorical_accuracy: 0.9100
Epoch 30/62
21/21 [==============================] - ETA: 0s - loss: 0.0372 - categorical_accuracy: 0.9985
Epoch 00030: saving model to model_2021-06-2711_07_38.791020/model-00030-0.03724-0.99849-0.25441-0.92000.h5
21/21 [==============================] - 69s 3s/step - loss: 0.0372 - categorical_accuracy: 0.9985 - val_loss: 0.2544 - val_categorical_accuracy: 0.9200
Epoch 31/62
21/21 [==============================] - ETA: 0s - loss: 0.0554 - categorical_accuracy: 0.9849
Epoch 00031: saving model to model_2021-06-2711_07_38.791020/model-00031-0.05535-0.98492-0.44432-0.88000.h5
21/21 [==============================] - 71s 3s/step - loss: 0.0554 - categorical_accuracy: 0.9849 - val_loss: 0.4443 - val_categorical_accuracy: 0.8800
Epoch 32/62
21/21 [==============================] - ETA: 0s - loss: 0.0475 - categorical_accuracy: 0.9804
Epoch 00032: saving model to model_2021-06-2711_07_38.791020/model-00032-0.04749-0.98039-0.32200-0.90000.h5
21/21 [==============================] - 67s 3s/step - loss: 0.0475 - categorical_accuracy: 0.9804 - val_loss: 0.3220 - val_categorical_accuracy: 0.9000
Epoch 33/62
21/21 [==============================] - ETA: 0s - loss: 0.0504 - categorical_accuracy: 0.9879
Epoch 00033: saving model to model_2021-06-2711_07_38.791020/model-00033-0.05043-0.98793-0.32458-0.90000.h5
21/21 [==============================] - 69s 3s/step - loss: 0.0504 - categorical_accuracy: 0.9879 - val_loss: 0.3246 - val_categorical_accuracy: 0.9000
Epoch 34/62
21/21 [==============================] - ETA: 0s - loss: 0.0554 - categorical_accuracy: 0.9834
Epoch 00034: saving model to model_2021-06-2711_07_38.791020/model-00034-0.05539-0.98341-0.33143-0.92000.h5
21/21 [==============================] - 71s 3s/step - loss: 0.0554 - categorical_accuracy: 0.9834 - val_loss: 0.3314 - val_categorical_accuracy: 0.9200
Epoch 35/62
21/21 [==============================] - ETA: 0s - loss: 0.0382 - categorical_accuracy: 0.9940
Epoch 00035: saving model to model_2021-06-2711_07_38.791020/model-00035-0.03825-0.99397-0.22969-0.92000.h5
21/21 [==============================] - 70s 3s/step - loss: 0.0382 - categorical_accuracy: 0.9940 - val_loss: 0.2297 - val_categorical_accuracy: 0.9200
Epoch 36/62
21/21 [==============================] - ETA: 0s - loss: 0.0575 - categorical_accuracy: 0.9819
Epoch 00036: saving model to model_2021-06-2711_07_38.791020/model-00036-0.05748-0.98190-0.33281-0.91000.h5
21/21 [==============================] - 68s 3s/step - loss: 0.0575 - categorical_accuracy: 0.9819 - val_loss: 0.3328 - val_categorical_accuracy: 0.9100
Epoch 37/62
21/21 [==============================] - ETA: 0s - loss: 0.0530 - categorical_accuracy: 0.9894
Epoch 00037: saving model to model_2021-06-2711_07_38.791020/model-00037-0.05304-0.98944-0.35337-0.90000.h5
21/21 [==============================] - 70s 3s/step - loss: 0.0530 - categorical_accuracy: 0.9894 - val_loss: 0.3534 - val_categorical_accuracy: 0.9000
Epoch 38/62
21/21 [==============================] - ETA: 0s - loss: 0.0530 - categorical_accuracy: 0.9804
Epoch 00038: saving model to model_2021-06-2711_07_38.791020/model-00038-0.05304-0.98039-0.37319-0.89000.h5
21/21 [==============================] - 71s 3s/step - loss: 0.0530 - categorical_accuracy: 0.9804 - val_loss: 0.3732 - val_categorical_accuracy: 0.8900
Epoch 39/62
21/21 [==============================] - ETA: 0s - loss: 0.0403 - categorical_accuracy: 0.9864
Epoch 00039: saving model to model_2021-06-2711_07_38.791020/model-00039-0.04028-0.98643-0.40226-0.89000.h5
21/21 [==============================] - 72s 3s/step - loss: 0.0403 - categorical_accuracy: 0.9864 - val_loss: 0.4023 - val_categorical_accuracy: 0.8900
Epoch 40/62
21/21 [==============================] - ETA: 0s - loss: 0.0488 - categorical_accuracy: 0.9879
Epoch 00040: saving model to model_2021-06-2711_07_38.791020/model-00040-0.04883-0.98793-0.32490-0.89000.h5
Epoch 00040: ReduceLROnPlateau reducing learning rate to 8.999999408842996e-07.
21/21 [==============================] - 69s 3s/step - loss: 0.0488 - categorical_accuracy: 0.9879 - val_loss: 0.3249 - val_categorical_accuracy: 0.8900
Epoch 41/62
21/21 [==============================] - ETA: 0s - loss: 0.0406 - categorical_accuracy: 0.9910
Epoch 00041: saving model to model_2021-06-2711_07_38.791020/model-00041-0.04058-0.99095-0.34474-0.88000.h5
21/21 [==============================] - 71s 3s/step - loss: 0.0406 - categorical_accuracy: 0.9910 - val_loss: 0.3447 - val_categorical_accuracy: 0.8800
Epoch 42/62
21/21 [==============================] - ETA: 0s - loss: 0.0299 - categorical_accuracy: 0.9894
Epoch 00042: saving model to model_2021-06-2711_07_38.791020/model-00042-0.02989-0.98944-0.28578-0.90000.h5
21/21 [==============================] - 70s 3s/step - loss: 0.0299 - categorical_accuracy: 0.9894 - val_loss: 0.2858 - val_categorical_accuracy: 0.9000
Epoch 43/62
21/21 [==============================] - ETA: 0s - loss: 0.0315 - categorical_accuracy: 0.9925
Epoch 00043: saving model to model_2021-06-2711_07_38.791020/model-00043-0.03147-0.99246-0.22249-0.92000.h5
21/21 [==============================] - 71s 3s/step - loss: 0.0315 - categorical_accuracy: 0.9925 - val_loss: 0.2225 - val_categorical_accuracy: 0.9200
Epoch 44/62
21/21 [==============================] - ETA: 0s - loss: 0.0532 - categorical_accuracy: 0.9804
Epoch 00044: saving model to model_2021-06-2711_07_38.791020/model-00044-0.05321-0.98039-0.33795-0.90000.h5
21/21 [==============================] - 69s 3s/step - loss: 0.0532 - categorical_accuracy: 0.9804 - val_loss: 0.3379 - val_categorical_accuracy: 0.9000
Epoch 45/62
21/21 [==============================] - ETA: 0s - loss: 0.0300 - categorical_accuracy: 0.9955
Epoch 00045: saving model to model_2021-06-2711_07_38.791020/model-00045-0.03000-0.99548-0.33273-0.90000.h5
21/21 [==============================] - 70s 3s/step - loss: 0.0300 - categorical_accuracy: 0.9955 - val_loss: 0.3327 - val_categorical_accuracy: 0.9000
Epoch 46/62
21/21 [==============================] - ETA: 0s - loss: 0.0536 - categorical_accuracy: 0.9834
Epoch 00046: saving model to model_2021-06-2711_07_38.791020/model-00046-0.05357-0.98341-0.38007-0.91000.h5
21/21 [==============================] - 74s 4s/step - loss: 0.0536 - categorical_accuracy: 0.9834 - val_loss: 0.3801 - val_categorical_accuracy: 0.9100
Epoch 47/62
21/21 [==============================] - ETA: 0s - loss: 0.0408 - categorical_accuracy: 0.9894
Epoch 00047: saving model to model_2021-06-2711_07_38.791020/model-00047-0.04085-0.98944-0.27676-0.90000.h5
21/21 [==============================] - 71s 3s/step - loss: 0.0408 - categorical_accuracy: 0.9894 - val_loss: 0.2768 - val_categorical_accuracy: 0.9000
Epoch 48/62
21/21 [==============================] - ETA: 0s - loss: 0.0376 - categorical_accuracy: 0.9864
Epoch 00048: saving model to model_2021-06-2711_07_38.791020/model-00048-0.03759-0.98643-0.32985-0.90000.h5
Epoch 00048: ReduceLROnPlateau reducing learning rate to 8.99999918146932e-09.
21/21 [==============================] - 71s 3s/step - loss: 0.0376 - categorical_accuracy: 0.9864 - val_loss: 0.3299 - val_categorical_accuracy: 0.9000
Epoch 49/62
21/21 [==============================] - ETA: 0s - loss: 0.0507 - categorical_accuracy: 0.9834
Epoch 00049: saving model to model_2021-06-2711_07_38.791020/model-00049-0.05072-0.98341-0.33805-0.88000.h5
21/21 [==============================] - 73s 3s/step - loss: 0.0507 - categorical_accuracy: 0.9834 - val_loss: 0.3381 - val_categorical_accuracy: 0.8800
Epoch 50/62
21/21 [==============================] - ETA: 0s - loss: 0.0279 - categorical_accuracy: 0.9985
Epoch 00050: saving model to model_2021-06-2711_07_38.791020/model-00050-0.02792-0.99849-0.21931-0.92000.h5
21/21 [==============================] - 73s 3s/step - loss: 0.0279 - categorical_accuracy: 0.9985 - val_loss: 0.2193 - val_categorical_accuracy: 0.9200
Epoch 51/62
21/21 [==============================] - ETA: 0s - loss: 0.0368 - categorical_accuracy: 0.9925
Epoch 00051: saving model to model_2021-06-2711_07_38.791020/model-00051-0.03677-0.99246-0.30997-0.89000.h5
21/21 [==============================] - 72s 3s/step - loss: 0.0368 - categorical_accuracy: 0.9925 - val_loss: 0.3100 - val_categorical_accuracy: 0.8900
Epoch 52/62
21/21 [==============================] - ETA: 0s - loss: 0.0377 - categorical_accuracy: 0.9925
Epoch 00052: saving model to model_2021-06-2711_07_38.791020/model-00052-0.03765-0.99246-0.33406-0.89000.h5
21/21 [==============================] - 69s 3s/step - loss: 0.0377 - categorical_accuracy: 0.9925 - val_loss: 0.3341 - val_categorical_accuracy: 0.8900
Epoch 53/62
21/21 [==============================] - ETA: 0s - loss: 0.0551 - categorical_accuracy: 0.9804
Epoch 00053: saving model to model_2021-06-2711_07_38.791020/model-00053-0.05512-0.98039-0.38544-0.90000.h5
21/21 [==============================] - 72s 3s/step - loss: 0.0551 - categorical_accuracy: 0.9804 - val_loss: 0.3854 - val_categorical_accuracy: 0.9000
Epoch 54/62
21/21 [==============================] - ETA: 0s - loss: 0.0476 - categorical_accuracy: 0.9819
Epoch 00054: saving model to model_2021-06-2711_07_38.791020/model-00054-0.04758-0.98190-0.38058-0.90000.h5
21/21 [==============================] - 72s 3s/step - loss: 0.0476 - categorical_accuracy: 0.9819 - val_loss: 0.3806 - val_categorical_accuracy: 0.9000
Epoch 55/62
21/21 [==============================] - ETA: 0s - loss: 0.0322 - categorical_accuracy: 0.9955
Epoch 00055: saving model to model_2021-06-2711_07_38.791020/model-00055-0.03224-0.99548-0.31970-0.88000.h5
21/21 [==============================] - 73s 3s/step - loss: 0.0322 - categorical_accuracy: 0.9955 - val_loss: 0.3197 - val_categorical_accuracy: 0.8800
Epoch 56/62
21/21 [==============================] - ETA: 0s - loss: 0.0343 - categorical_accuracy: 0.9910
Epoch 00056: saving model to model_2021-06-2711_07_38.791020/model-00056-0.03427-0.99095-0.33414-0.90000.h5
Epoch 00056: ReduceLROnPlateau reducing learning rate to 8.999998968306499e-11.
21/21 [==============================] - 70s 3s/step - loss: 0.0343 - categorical_accuracy: 0.9910 - val_loss: 0.3341 - val_categorical_accuracy: 0.9000
Epoch 57/62
21/21 [==============================] - ETA: 0s - loss: 0.0320 - categorical_accuracy: 0.9940
Epoch 00057: saving model to model_2021-06-2711_07_38.791020/model-00057-0.03203-0.99397-0.26992-0.91000.h5
21/21 [==============================] - 72s 3s/step - loss: 0.0320 - categorical_accuracy: 0.9940 - val_loss: 0.2699 - val_categorical_accuracy: 0.9100
Epoch 58/62
21/21 [==============================] - ETA: 0s - loss: 0.0543 - categorical_accuracy: 0.9834
Epoch 00058: saving model to model_2021-06-2711_07_38.791020/model-00058-0.05431-0.98341-0.28473-0.93000.h5
21/21 [==============================] - 74s 4s/step - loss: 0.0543 - categorical_accuracy: 0.9834 - val_loss: 0.2847 - val_categorical_accuracy: 0.9300
Epoch 59/62
21/21 [==============================] - ETA: 0s - loss: 0.0337 - categorical_accuracy: 0.9955
Epoch 00059: saving model to model_2021-06-2711_07_38.791020/model-00059-0.03368-0.99548-0.32889-0.90000.h5
21/21 [==============================] - 74s 4s/step - loss: 0.0337 - categorical_accuracy: 0.9955 - val_loss: 0.3289 - val_categorical_accuracy: 0.9000
Epoch 60/62
21/21 [==============================] - ETA: 0s - loss: 0.0418 - categorical_accuracy: 0.9849
Epoch 00060: saving model to model_2021-06-2711_07_38.791020/model-00060-0.04184-0.98492-0.32163-0.90000.h5
21/21 [==============================] - 74s 4s/step - loss: 0.0418 - categorical_accuracy: 0.9849 - val_loss: 0.3216 - val_categorical_accuracy: 0.9000
Epoch 61/62
21/21 [==============================] - ETA: 0s - loss: 0.0321 - categorical_accuracy: 0.9925
Epoch 00061: saving model to model_2021-06-2711_07_38.791020/model-00061-0.03213-0.99246-0.34255-0.90000.h5
21/21 [==============================] - 77s 4s/step - loss: 0.0321 - categorical_accuracy: 0.9925 - val_loss: 0.3426 - val_categorical_accuracy: 0.9000
Epoch 62/62
21/21 [==============================] - ETA: 0s - loss: 0.0452 - categorical_accuracy: 0.9819
Epoch 00062: saving model to model_2021-06-2711_07_38.791020/model-00062-0.04522-0.98190-0.14211-0.94000.h5
21/21 [==============================] - 73s 3s/step - loss: 0.0452 - categorical_accuracy: 0.9819 - val_loss: 0.1421 - val_categorical_accuracy: 0.9400
import matplotlib.pyplot as plt
plot(history)