The code includes 3 main parts as below:

Split the image
Clean the images
Learning and prediction

1. Split the images

This code will: 1. Upload the original (.h5) dataset, 2. Store the .h5 file, to 3 variables, X_train, X_test, and y_train. 2. Convert each image in X_train and X_test(6464) to 5, 1414 images. 3. Flatten the 560005 y_train to 28000. 4. Save 3 outputs including X_train (2800001414), X_test (7000014*14), and y_train (280000,)

Import libraries:

import numpy as np from numpy import asarray, array from numpy import savetxt, save, load import pandas as pd import pdb import time from sklearn.datasets import load_digits, fetch_openml from sklearn.linear_model import Ridge from sklearn.metrics import classification_report, confusion_matrix, plot_confusion_matrix import matplotlib.pyplot as plt from sklearn.model_selection import train_test_split, validation_curve, cross_val_score from sklearn.utils import check_random_state from IPython.core.debugger import set_trace import warnings warnings.filterwarnings('ignore') %matplotlib inline import h5py from matplotlib.image import imread from PIL import Image as im from scipy import ndimage from skimage import filters import math import tensorflow as tf import matplotlib.pyplot as plt from PIL import Image

Open the original dataset from google drive:

##### open the file and extract to the test and train arrays ##### im_shape_org = (64, 64) ds = [] with h5py.File('/content/drive/MyDrive/Colab Notebooks/Datasets/MNIST_synthetic.h5', 'r') as f: for key in f.keys(): ds.append(f[key][()]) print(key) # using this command to see the keys in the dataset ds_X_tst_org, ds_X_trn_org, ds_y_trn_org = ds[0], ds[1], ds[2] ds_tst_smpl_nr = ds_X_tst_org.shape[0] ds_trn_smpl_nr = ds_X_trn_org.shape[0] print(ds_X_tst_org.shape, ds_X_trn_org.shape, ds_y_trn_org.shape) #check the shape of dataset

Predefined functions:

In this section, we defined the required function for to split the images. The main steps are as below:

Crop the image.
List item Split the image based on the optimum digit width (dgt_width).
Put the slices to a new dataset.

def image_split_object(ds_arr, digit_width = 12, display_samples=False): ##### initialize some variables ##### im_shape_obj = (14,14) smpl_nr = ds_arr.shape[0] cls_nr = 5 critical_ind = np.array([]) img_index = 0 dgt_width = digit_width ####################################### ##### Change all the elements to a zero array of shape im_shape ##### img_zero = np.zeros(im_shape_obj) ds_new = np.full([smpl_nr, cls_nr], None) for rw in range(smpl_nr): for col in range(cls_nr): ds_new[rw,col] = img_zero ####################################################################### ##### split each image to the objects and creat a new dataset ##### for img_index in range(smpl_nr): #range(smpl_nr): ## upload the image from dataset ## img_curr = np. array(ds_arr[img_index]) img_curr = img_curr.reshape(im_shape_org) ## plot the current sample ## if img_index == 0 and display_samples==True: # just the 1st image for test print("digits width:", dgt_width) plt.imshow(img_curr) plt.title('The first image to split.', fontsize=8) plt.show() ############################### ##################################### ## crop the image and remove zero margin ## zero_cols = np.argwhere(np.all(img_curr[:, :] == 0, axis=0)) nzero_cols = np.argwhere(np.any(img_curr[:, :] != 0, axis=0)) zero_rows = np.argwhere(np.all(img_curr[:, :] == 0, axis=1)) nzero_rows = np.argwhere(np.any(img_curr[:, :] != 0, axis=1)) img_curr = img_curr[nzero_rows.min()-1:nzero_rows.max()+1, nzero_cols.min()-1:nzero_cols.min()-1+(cls_nr*dgt_width)] ## plot the current sample ## if img_index == 0 and display_samples==True: # just the 1st image for test plt.imshow(img_curr) plt.title('The image after crop.', fontsize=8) plt.show() ############################### ############################################# ## extract the image size and possible number of objects ## frm_size = img_curr.shape[1] dgt_nr = math.ceil(frm_size/dgt_width) ############################################################# ## insert the objects in the new dataset and remove redundant objects ## obj_ind = 0 obj_last_ind = 5 for j in range(dgt_nr): if j == 0: start_col = 0 else: start_col = (j*dgt_width)-2 end_col = start_col + dgt_width tmp_im = img_curr[0:img_curr.shape[0], start_col:end_col] ## plot the current sample ## if img_index == 0 and display_samples==True: # just the 1st image for test plt.imshow(tmp_im) plt.title('Digit after splitting', fontsize=8) plt.show() ############################### if np.count_nonzero(tmp_im) > 0.1*tmp_im.size: # remove redundant objects with mostly (0.1*tmp_im.size) zero pixels ds_tmp = np.zeros(im_shape_obj) ds_tmp[:tmp_im.shape[0],:tmp_im.shape[1]] = tmp_im #[:tmp_im.shape[0],:tmp_im.shape[1]] ds_new[img_index, obj_ind] = ds_tmp obj_ind += 1 ########################################################################## ##### compare the num of found objects to y_trn ##### # ind_cls_10 = np.where(ds_y_trn_org[img_index]==10) # y_trim = np.delete(ds_y_trn_org[img_index], ind_cls_10) # if obj_ind != y_trim.shape[0]: # np.append(critical_ind, img_index) # critical_ind.append(img_index) # print("size diff: NOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO") ####################################################### return ds_new, critical_ind def grid_show(images_array, rows=3, columns=5, first_ind = 0): ##### display the loaded file to check the new dataset ##### num_row = rows num_col = columns num_of_images = num_row*num_col # data.shape[0] # plot images fig, axes = plt.subplots(num_row, num_col, figsize=(10,10)) for i in range(num_of_images): image = images_array[i+first_ind] ax = axes[i//num_col, i%num_col] ax.imshow(image, cmap='gray') # ax.set_title('Label: {}'.format(ds_y_trn_fltn[i])) plt.tight_layout() plt.show() ##############################################################

##### split the objects ##### ds_X_tst_splt, X_tst_crit_ind = image_split_object(ds_X_tst_org, digit_width=12) ds_X_tst_splt_fltn = np.ndarray.flatten(ds_X_tst_splt) ds_X_trn_splt, X_trn_crit_ind = image_split_object(ds_X_trn_org, digit_width=12) ds_X_trn_splt_fltn = np.ndarray.flatten(ds_X_trn_splt) ds_y_trn_fltn = np.ndarray.flatten(ds_y_trn_org) ###############################

Tunning the digits width and visualizing the split process:

The following part is just to display the process of the splitting for digits width of 11, 12, and 13:

for i in range(3): image_split_object(ds_X_trn_org[0:1], digit_width=11+i, display_samples=True)

Finalizing the images:

Increasing the size of each image from 14x14, to 18x18

##### Pad the images and change from 14*14 to 18*18 ##### if ds_X_tst_splt_fltn[0].shape == (14,14): for p in range(ds_X_trn_splt_fltn.shape[0]): ds_X_trn_splt_fltn[p] = np.pad(ds_X_trn_splt_fltn[p], (2,2), 'constant', constant_values = 0) for p in range(ds_X_tst_splt_fltn.shape[0]): ds_X_tst_splt_fltn[p] = np.pad(ds_X_tst_splt_fltn[p], (2,2), 'constant', constant_values = 0) ###########################################################

Converting to array type:

##### do the conversion to convert from (x,) to (x,y)!!! ##### ds_X_tst_splt_fltn = np.asarray(list(ds_X_tst_splt_fltn)) ds_X_trn_splt_fltn = np.asarray(list(ds_X_trn_splt_fltn)) ################################################################

Save the new datasets:

Now we have three new datasets that we are going to save them as .npy files, in the allocated directory:

##### save the new datasets to files ##### from numpy import save, load save('/content/drive/MyDrive/Colab Notebooks/Datasets/ds_X_tst_splt_fltn-v02.npy', ds_X_tst_splt_fltn) save('/content/drive/MyDrive/Colab Notebooks/Datasets/ds_X_trn_splt_fltn-v02.npy', ds_X_trn_splt_fltn) save('/content/drive/MyDrive/Colab Notebooks/Datasets/ds_y_trn_fltn-v02.npy', ds_y_trn_fltn) ############################################

Validation:

In this section we test what we did in previouse section.

Loading datasets:

##### load the .npy file ##### data_trn = load('/content/drive/MyDrive/Colab Notebooks/Datasets/ds_X_trn_splt_fltn-v02.npy', allow_pickle=True) data_tst = load('/content/drive/MyDrive/Colab Notebooks/Datasets/ds_X_tst_splt_fltn-v02.npy', allow_pickle=True) print("Training data shape:", data_trn.shape) print("Test data shape:", data_tst.shape) ################################

Display the images in the new dataset:

##### display the loaded file to check the new dataset ##### grid_show(data_trn) ##############################################################

2. Cleaning the images:

In the last section we splited the images and created new dataset. But some images, beside the main digit in the image, include some small objects that we are going to detect and remove them in this part.

This code will import the X_train, and X_test, and do the cleaning as follow: 1. Upload the X_train (280000,14,14) and X_test (56000,14,14). 2. Do the blur and mask methods on each image in datasets. 3. Extract the number of objects in each image. Each image should include just one object. 4. Remove the small objects in multi object images. 5. Save the new dataset.

Import libraries

import numpy as np from numpy import asarray, array from numpy import savetxt, save, load import pandas as pd import tensorflow as tf import pdb import time from sklearn.datasets import load_digits, fetch_openml from sklearn.linear_model import Ridge from sklearn.metrics import classification_report, confusion_matrix, plot_confusion_matrix import matplotlib.pyplot as plt from sklearn.model_selection import train_test_split, validation_curve, cross_val_score from sklearn.naive_bayes import GaussianNB from sklearn.neighbors import KNeighborsClassifier from sklearn.tree import DecisionTreeClassifier from sklearn.utils import check_random_state from IPython.core.debugger import set_trace import warnings warnings.filterwarnings('ignore') %matplotlib inline import h5py from matplotlib.image import imread from PIL import Image from scipy import ndimage from skimage import filters

Loading datasets

The new datasets that generated in previous section, are loaded again to be manipulated!

##### load the .npy file ##### X_trn_arr = load('/content/drive/MyDrive/Colab Notebooks/Datasets/ds_X_trn_splt_fltn-v02.npy', allow_pickle=True) X_tst_arr = load('/content/drive/MyDrive/Colab Notebooks/Datasets/ds_X_tst_splt_fltn-v02.npy', allow_pickle=True) ################################

Predefined functions:

Here we defined the functions to do the blur and mask on each image, and also the function to display the images.

# smooth the image (to remove small objects) def image_blur(im_inp, blur_radius, disp_image=False): blr_rds = blur_radius im_blr = ndimage.gaussian_filter(im_inp, blr_rds) if disp_image == True: plt.imshow(im_blr) plt.show() return im_blr def image_mask(im_input, threshold): if threshold == 1: thrsh = filters.threshold_otsu(im_input) elif threshold == 2: thrsh = filters.threshold_isodata(im_input) elif threshold == 3: thrsh = filters.threshold_sauvola(im_input) else: thrsh = threshold ### im_mask = im_input > threshold ### return im_mask

Tunning the blur radiation:

In the following section the proper value for blur radiation tunned.

ds = X_trn_arr[36:37] ds_tmp = ds ds_tmp_cln = ds_tmp # initialize the output dataset dgt_size = (18,18) im_mult_obj = np.zeros(ds_tmp.shape[0]) # initialize an array to store the indeces of multi object images. for i in range(ds_tmp.shape[0]): if np.any(ds_tmp[i] != 0): # empty images can not be blured print(i) plt.imshow(ds_tmp[i]) plt.title('The original image after splitting.', fontsize=8) plt.show() for j in [0.2, 0.6, 1]: ds_tmp_blr = image_blur(ds_tmp[i], j) plt.imshow(ds_tmp_blr) plt.title("The image blur radiation:%1.1f"%j , fontsize=8) plt.show() ds_tmp_blr_msk = image_mask(ds_tmp_blr, 1) plt.imshow(ds_tmp_blr_msk) plt.title('The image masked', fontsize=8) plt.show()

Multi-object detection:

In this part each image will be blured, then masked, and finally, labeled to provide the proper situation for object detection.

In the next step, multi object images will be analysed and objects with small size will be deleted.

ds_num = 0 ds_cln = [0, 0] for ds in [X_tst_arr, X_trn_arr]: ds_tmp = ds ds_tmp_cln = ds_tmp # initialize the output dataset dgt_size = (18,18) im_mult_obj = np.zeros(ds_tmp.shape[0]) # initialize an array to store the indeces of multi object images. for i in range(ds_tmp.shape[0]): if np.any(ds_tmp[i] != 0): # empty images can not be blured ds_tmp_blr = image_blur(ds_tmp[i], 0.2) ds_tmp_blr_msk = image_mask(ds_tmp_blr, 1) labeling_structure = [[1, 1, 1],[1, 1, 1],[1, 1, 1]] ds_tmp_lbld, nr_objects = ndimage.label(ds_tmp_blr_msk, structure=labeling_structure) obj_slice = (ndimage.find_objects(ds_tmp_lbld)) if nr_objects>1: for z in range(nr_objects): obj_tmp = ds_tmp[i][obj_slice[z]] obj_size_tmp = np.count_nonzero(obj_tmp) if obj_size_tmp<10: ds_tmp_cln[i][obj_slice[z]] = 0 ds_cln[ds_num] = ds_tmp_cln ds_num +=1

Save the new dataset:

When the cleaning is done, we save the new dataset and then we can use it for training and learning process.

# save('ds_X_trn_cln.npy', ds_tmp_cln) save('/content/drive/MyDrive/Colab Notebooks/Datasets/ds_X_tst_cln-v02.npy', ds_cln[0]) save('/content/drive/MyDrive/Colab Notebooks/Datasets/ds_X_trn_cln-v02.npy', ds_cln[1])

Validation phase:

This section is for a primary validation and visualizing the clean dataset. The main validation is done in training step.

Load the dataset:

X_trn_splt_org = load('/content/drive/MyDrive/Colab Notebooks/Datasets/ds_X_trn_splt_fltn-v02.npy') X_tst_splt_org = load('/content/drive/MyDrive/Colab Notebooks/Datasets/ds_X_tst_splt_fltn-v02.npy') X_tst_splt_cln = load('/content/drive/MyDrive/Colab Notebooks/Datasets/ds_X_tst_cln-v02.npy') X_trn_splt_cln = load('/content/drive/MyDrive/Colab Notebooks/Datasets/ds_X_trn_cln-v02.npy') y_trn_fltn = load('/content/drive/MyDrive/Colab Notebooks/Datasets/ds_y_trn_fltn-v02.npy')

Display the images:

In this section we will display and compare the result of cleaning process.

##### display the loaded file to check the new dataset ##### plt.imshow(X_trn_splt_org[1]) plt.show() plt.imshow(X_trn_splt_cln[1]) plt.show() ##############################################################

Building and optimizing a model

Import libraries

import tensorflow as tf import numpy as np from numpy import mean from numpy import std from numpy import load from matplotlib import pyplot from sklearn.model_selection import KFold, train_test_split from keras.datasets import mnist from keras.utils import to_categorical from keras.models import Sequential from keras.layers import Conv2D from keras.layers import MaxPooling2D from keras.layers import Dense from keras.layers import Flatten from keras.layers import Dropout from keras.optimizers import Adam, SGD from keras.constraints import maxnorm from keras.layers import BatchNormalization from keras.callbacks import ModelCheckpoint, ReduceLROnPlateau from keras.preprocessing.image import ImageDataGenerator from sklearn.model_selection import GridSearchCV from keras.wrappers.scikit_learn import KerasClassifier from keras.utils import plot_model from google.colab import drive drive.mount('/content/gdrive/', force_remount=True)

load the preprocessed datasets

def load_our_dataset(): X_train = load('gdrive/MyDrive/ds_X_trn_cln.npy', allow_pickle=True) X_train= X_train.reshape((X_train.shape[0], 14, 14, 1)) X_train = tf.keras.utils.normalize(X_train, axis=1) X_test = load('gdrive/MyDrive/ds_X_tst_cln.npy', allow_pickle=True) X_test= X_test.reshape((X_test.shape[0], 14, 14, 1)) X_test = tf.keras.utils.normalize(X_test, axis=1) Y_train = load('gdrive/MyDrive/ds_y_trn_fltn.npy', allow_pickle=True) Y_train = to_categorical(Y_train) return X_train, Y_train, X_test

building our model

2 convolution layers + pooling + 2 dense layers + 1 final dense layer

def build_cnn(lrx=0.001, momentumx=0.9, epochsx=22): model = Sequential() model.add(Conv2D(32, (3, 3), activation='relu', padding='same', kernel_initializer='lecun_uniform', input_shape=(14, 14, 1))) model.add(BatchNormalization()) model.add(Conv2D(64, (3, 3), activation='relu', padding='same', kernel_initializer='lecun_uniform')) model.add(MaxPooling2D((2, 2))) model.add(Flatten()) model.add(Dropout(0.2)) model.add(Dense(120, activation='relu', kernel_initializer='lecun_uniform')) model.add(Dropout(0.16)) model.add(Dense(55, activation='relu', kernel_initializer='lecun_uniform')) model.add(Dense(11, activation='softmax')) opt = Adam(lr=lrx) model.compile(optimizer=opt ,loss='categorical_crossentropy', metrics=['accuracy']) return model

Evaluate the model with 5-fold cross-validation

The best model (with least validation_loss) will be saved, moreover, learning rate reduction was used

def cross_validation(dataX, dataY, n_folds=5, epochsx=22, batch_sizex=32, lrx=0.01, momentumx=0.9): scores, histories = list(), list() kfold = KFold(n_folds, shuffle=True, random_state=1) for train_index, test_index in kfold.split(dataX): cnn = build_cnn(lrx, momentumx, epochsx) X_train, Y_train, X_test, Y_test = dataX[train_index], dataY[train_index], dataX[test_index], dataY[test_index] learning_rate_reduction = ReduceLROnPlateau(monitor='val_loss', patience=3, verbose=1, factor=0.5, min_lr=0.000001) history = cnn.fit(X_train, Y_train , epochs = epochsx, validation_data= (X_test, Y_test),callbacks=[checkpoint, learning_rate_reduction], verbose=1) _, acc = cnn.evaluate(X_test, Y_test, verbose=0) print('> %.3f' % (acc * 100.0)) scores.append(acc) histories.append(history) return scores, histories

Ploting the learning curves of the model(s)

def cnn_performance(scores, histories): pyplot.figure(figsize=(10,8)) for i in range(len(histories)): pyplot.subplot(2, 1, 1) pyplot.title('Cross Entropy Loss') pyplot.plot(histories[i].history['loss'], color='blue', label='train') pyplot.plot(histories[i].history['val_loss'], color='orange', label='test') pyplot.subplot(2, 1, 2) pyplot.title('Classification Accuracy') pyplot.plot(histories[i].history['accuracy'], color='blue', label='train') pyplot.plot(histories[i].history['val_accuracy'], color='orange', label='test') pyplot.show() print('Accuracy: mean=%.3f std=%.3f, n=%d' % (mean(scores)*100, std(scores)*100, len(scores))) pyplot.boxplot(scores) pyplot.show()

Setting up running

Here we can run everything as to test the architecture of our model as well as choosing hyperparameters

def run_test_cnn(): X_train, Y_train, X_test = load_our_dataset() scores, histories = cross_validation(X_train, Y_train,lrx=0.001, epochsx=30, batch_sizex=60) cnn_performance(scores, histories)

Saving the best model

based on best validation loss

checkpoint = ModelCheckpoint('best_model_improved.h5', # model filename monitor='val_loss', # quantity to monitor verbose=0, # verbosity - 0 or 1 save_best_only= True, # The latest best model will not be overwritten mode='auto')

run_test_cnn()

Using our best model for predictions

def cnn_predict(): _, _, X_test = load_our_dataset() cnn = tf.keras.models.load_model('best_model_improved.h5') predictions=np.argmax(cnn.predict([X_test]), axis=-1) predictions= concat_output(predictions) np.savetxt("final_predictions1.csv", predictions, delimiter=",", fmt='%s')

cnn_predict()

Epoch 1/30 7000/7000 ============================== - 22s 3ms/step - loss: 0.1247 - accuracy: 0.9613 - val_loss: 0.0394 - val_accuracy: 0.9873 Epoch 2/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0543 - accuracy: 0.9842 - val_loss: 0.0299 - val_accuracy: 0.9913 Epoch 3/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0407 - accuracy: 0.9879 - val_loss: 0.0260 - val_accuracy: 0.9923 Epoch 4/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0323 - accuracy: 0.9903 - val_loss: 0.0294 - val_accuracy: 0.9914 Epoch 5/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0291 - accuracy: 0.9914 - val_loss: 0.0221 - val_accuracy: 0.9937 Epoch 6/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0260 - accuracy: 0.9921 - val_loss: 0.0240 - val_accuracy: 0.9937 Epoch 7/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0235 - accuracy: 0.9929 - val_loss: 0.0266 - val_accuracy: 0.9931 Epoch 8/30 6988/7000 ============================>. - ETA: 0s - loss: 0.0224 - accuracy: 0.9934 Epoch 00008: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0224 - accuracy: 0.9934 - val_loss: 0.0227 - val_accuracy: 0.9935 Epoch 9/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0125 - accuracy: 0.9961 - val_loss: 0.0173 - val_accuracy: 0.9956 Epoch 10/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0101 - accuracy: 0.9968 - val_loss: 0.0181 - val_accuracy: 0.9954 Epoch 11/30 7000/7000 ============================== - 23s 3ms/step - loss: 0.0103 - accuracy: 0.9969 - val_loss: 0.0172 - val_accuracy: 0.9957 Epoch 12/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0091 - accuracy: 0.9973 - val_loss: 0.0164 - val_accuracy: 0.9959 Epoch 13/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0087 - accuracy: 0.9973 - val_loss: 0.0177 - val_accuracy: 0.9955 Epoch 14/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0084 - accuracy: 0.9975 - val_loss: 0.0172 - val_accuracy: 0.9958 Epoch 15/30 6995/7000 ============================>. - ETA: 0s - loss: 0.0078 - accuracy: 0.9977 Epoch 00015: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0078 - accuracy: 0.9977 - val_loss: 0.0197 - val_accuracy: 0.9956 Epoch 16/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0052 - accuracy: 0.9984 - val_loss: 0.0193 - val_accuracy: 0.9962 Epoch 17/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0042 - accuracy: 0.9986 - val_loss: 0.0180 - val_accuracy: 0.9965 Epoch 18/30 6984/7000 ============================>. - ETA: 0s - loss: 0.0039 - accuracy: 0.9988 Epoch 00018: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0039 - accuracy: 0.9988 - val_loss: 0.0195 - val_accuracy: 0.9965 Epoch 19/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0037 - accuracy: 0.9990 - val_loss: 0.0155 - val_accuracy: 0.9970 Epoch 20/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0027 - accuracy: 0.9991 - val_loss: 0.0164 - val_accuracy: 0.9968 Epoch 21/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0030 - accuracy: 0.9991 - val_loss: 0.0189 - val_accuracy: 0.9968 Epoch 22/30 6999/7000 ============================>. - ETA: 0s - loss: 0.0025 - accuracy: 0.9992 Epoch 00022: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0025 - accuracy: 0.9992 - val_loss: 0.0186 - val_accuracy: 0.9969 Epoch 23/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0023 - accuracy: 0.9992 - val_loss: 0.0177 - val_accuracy: 0.9968 Epoch 24/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0021 - accuracy: 0.9993 - val_loss: 0.0163 - val_accuracy: 0.9969 Epoch 25/30 6995/7000 ============================>. - ETA: 0s - loss: 0.0019 - accuracy: 0.9994 Epoch 00025: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0019 - accuracy: 0.9994 - val_loss: 0.0170 - val_accuracy: 0.9969 Epoch 26/30 7000/7000 ============================== - 23s 3ms/step - loss: 0.0019 - accuracy: 0.9995 - val_loss: 0.0169 - val_accuracy: 0.9969 Epoch 27/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0018 - accuracy: 0.9995 - val_loss: 0.0164 - val_accuracy: 0.9970 Epoch 28/30 6994/7000 ============================>. - ETA: 0s - loss: 0.0019 - accuracy: 0.9994 Epoch 00028: ReduceLROnPlateau reducing learning rate to 1.5625000742147677e-05. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0019 - accuracy: 0.9994 - val_loss: 0.0163 - val_accuracy: 0.9971 Epoch 29/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0017 - accuracy: 0.9996 - val_loss: 0.0175 - val_accuracy: 0.9971 Epoch 30/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0014 - accuracy: 0.9996 - val_loss: 0.0173 - val_accuracy: 0.9970

99.704 Epoch 1/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.1220 - accuracy: 0.9616 - val_loss: 0.0403 - val_accuracy: 0.9877 Epoch 2/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0552 - accuracy: 0.9836 - val_loss: 0.0302 - val_accuracy: 0.9905 Epoch 3/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0418 - accuracy: 0.9874 - val_loss: 0.0283 - val_accuracy: 0.9917 Epoch 4/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0347 - accuracy: 0.9896 - val_loss: 0.0286 - val_accuracy: 0.9919 Epoch 5/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0302 - accuracy: 0.9909 - val_loss: 0.0257 - val_accuracy: 0.9930 Epoch 6/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0255 - accuracy: 0.9921 - val_loss: 0.0289 - val_accuracy: 0.9926 Epoch 7/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0243 - accuracy: 0.9927 - val_loss: 0.0230 - val_accuracy: 0.9939 Epoch 8/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0233 - accuracy: 0.9932 - val_loss: 0.0257 - val_accuracy: 0.9937 Epoch 9/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0214 - accuracy: 0.9939 - val_loss: 0.0210 - val_accuracy: 0.9946 Epoch 10/30 7000/7000 ============================== - 22s 3ms/step - loss: 0.0194 - accuracy: 0.9942 - val_loss: 0.0218 - val_accuracy: 0.9947 Epoch 11/30 7000/7000 ============================== - 22s 3ms/step - loss: 0.0189 - accuracy: 0.9943 - val_loss: 0.0235 - val_accuracy: 0.9948 Epoch 12/30 6989/7000 ============================>. - ETA: 0s - loss: 0.0177 - accuracy: 0.9948 Epoch 00012: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0178 - accuracy: 0.9948 - val_loss: 0.0235 - val_accuracy: 0.9945 Epoch 13/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0098 - accuracy: 0.9969 - val_loss: 0.0167 - val_accuracy: 0.9962 Epoch 14/30 7000/7000 ============================== - 22s 3ms/step - loss: 0.0087 - accuracy: 0.9973 - val_loss: 0.0187 - val_accuracy: 0.9961 Epoch 15/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0079 - accuracy: 0.9976 - val_loss: 0.0183 - val_accuracy: 0.9963 Epoch 16/30 6991/7000 ============================>. - ETA: 0s - loss: 0.0073 - accuracy: 0.9978 Epoch 00016: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0073 - accuracy: 0.9978 - val_loss: 0.0225 - val_accuracy: 0.9959 Epoch 17/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0053 - accuracy: 0.9984 - val_loss: 0.0195 - val_accuracy: 0.9967 Epoch 18/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0046 - accuracy: 0.9986 - val_loss: 0.0203 - val_accuracy: 0.9962 Epoch 19/30 6982/7000 ============================>. - ETA: 0s - loss: 0.0044 - accuracy: 0.9987 Epoch 00019: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0044 - accuracy: 0.9987 - val_loss: 0.0182 - val_accuracy: 0.9968 Epoch 20/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0031 - accuracy: 0.9990 - val_loss: 0.0167 - val_accuracy: 0.9970 Epoch 21/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0026 - accuracy: 0.9992 - val_loss: 0.0181 - val_accuracy: 0.9969 Epoch 22/30 6989/7000 ============================>. - ETA: 0s - loss: 0.0023 - accuracy: 0.9993 Epoch 00022: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0023 - accuracy: 0.9993 - val_loss: 0.0188 - val_accuracy: 0.9969 Epoch 23/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0023 - accuracy: 0.9993 - val_loss: 0.0201 - val_accuracy: 0.9970 Epoch 24/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0023 - accuracy: 0.9992 - val_loss: 0.0188 - val_accuracy: 0.9970 Epoch 25/30 6990/7000 ============================>. - ETA: 0s - loss: 0.0020 - accuracy: 0.9994 Epoch 00025: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05. 7000/7000 ============================== - 23s 3ms/step - loss: 0.0020 - accuracy: 0.9994 - val_loss: 0.0187 - val_accuracy: 0.9970 Epoch 26/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0021 - accuracy: 0.9994 - val_loss: 0.0184 - val_accuracy: 0.9971 Epoch 27/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0017 - accuracy: 0.9994 - val_loss: 0.0194 - val_accuracy: 0.9971 Epoch 28/30 6988/7000 ============================>. - ETA: 0s - loss: 0.0017 - accuracy: 0.9995 Epoch 00028: ReduceLROnPlateau reducing learning rate to 1.5625000742147677e-05. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0017 - accuracy: 0.9995 - val_loss: 0.0194 - val_accuracy: 0.9970 Epoch 29/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0018 - accuracy: 0.9994 - val_loss: 0.0200 - val_accuracy: 0.9970 Epoch 30/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0017 - accuracy: 0.9995 - val_loss: 0.0197 - val_accuracy: 0.9970 99.700 Epoch 1/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.1237 - accuracy: 0.9621 - val_loss: 0.0470 - val_accuracy: 0.9860 Epoch 2/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0546 - accuracy: 0.9839 - val_loss: 0.0449 - val_accuracy: 0.9877 Epoch 3/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0420 - accuracy: 0.9875 - val_loss: 0.0255 - val_accuracy: 0.9923 Epoch 4/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0346 - accuracy: 0.9897 - val_loss: 0.0238 - val_accuracy: 0.9934 Epoch 5/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0300 - accuracy: 0.9910 - val_loss: 0.0346 - val_accuracy: 0.9912 Epoch 6/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0265 - accuracy: 0.9921 - val_loss: 0.0228 - val_accuracy: 0.9934 Epoch 7/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0237 - accuracy: 0.9930 - val_loss: 0.0289 - val_accuracy: 0.9927 Epoch 8/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0227 - accuracy: 0.9936 - val_loss: 0.0238 - val_accuracy: 0.9935 Epoch 9/30 7000/7000 ============================== - 22s 3ms/step - loss: 0.0209 - accuracy: 0.9937 - val_loss: 0.0225 - val_accuracy: 0.9944 Epoch 10/30 7000/7000 ============================== - 22s 3ms/step - loss: 0.0193 - accuracy: 0.9943 - val_loss: 0.0225 - val_accuracy: 0.9940 Epoch 11/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0179 - accuracy: 0.9947 - val_loss: 0.0207 - val_accuracy: 0.9943 Epoch 12/30 7000/7000 ============================== - 22s 3ms/step - loss: 0.0176 - accuracy: 0.9948 - val_loss: 0.0185 - val_accuracy: 0.9950 Epoch 13/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0161 - accuracy: 0.9953 - val_loss: 0.0202 - val_accuracy: 0.9950 Epoch 14/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0162 - accuracy: 0.9954 - val_loss: 0.0249 - val_accuracy: 0.9944 Epoch 15/30 6999/7000 ============================>. - ETA: 0s - loss: 0.0154 - accuracy: 0.9956 Epoch 00015: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0154 - accuracy: 0.9956 - val_loss: 0.0229 - val_accuracy: 0.9951 Epoch 16/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0092 - accuracy: 0.9972 - val_loss: 0.0177 - val_accuracy: 0.9960 Epoch 17/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0071 - accuracy: 0.9978 - val_loss: 0.0182 - val_accuracy: 0.9960 Epoch 18/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0073 - accuracy: 0.9980 - val_loss: 0.0194 - val_accuracy: 0.9956 Epoch 19/30 6990/7000 ============================>. - ETA: 0s - loss: 0.0067 - accuracy: 0.9981 Epoch 00019: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0067 - accuracy: 0.9981 - val_loss: 0.0221 - val_accuracy: 0.9956 Epoch 20/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0047 - accuracy: 0.9985 - val_loss: 0.0197 - val_accuracy: 0.9963 Epoch 21/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0040 - accuracy: 0.9988 - val_loss: 0.0169 - val_accuracy: 0.9965 Epoch 22/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0033 - accuracy: 0.9990 - val_loss: 0.0210 - val_accuracy: 0.9964 Epoch 23/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0039 - accuracy: 0.9989 - val_loss: 0.0182 - val_accuracy: 0.9967 Epoch 24/30 6992/7000 ============================>. - ETA: 0s - loss: 0.0035 - accuracy: 0.9990 Epoch 00024: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814. 7000/7000 ============================== - 23s 3ms/step - loss: 0.0035 - accuracy: 0.9990 - val_loss: 0.0193 - val_accuracy: 0.9966 Epoch 25/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0023 - accuracy: 0.9993 - val_loss: 0.0193 - val_accuracy: 0.9967 Epoch 26/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0025 - accuracy: 0.9992 - val_loss: 0.0199 - val_accuracy: 0.9965 Epoch 27/30 7000/7000 ============================== - ETA: 0s - loss: 0.0030 - accuracy: 0.9993 Epoch 00027: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0030 - accuracy: 0.9993 - val_loss: 0.0194 - val_accuracy: 0.9968 Epoch 28/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0019 - accuracy: 0.9994 - val_loss: 0.0210 - val_accuracy: 0.9967 Epoch 29/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0016 - accuracy: 0.9994 - val_loss: 0.0190 - val_accuracy: 0.9967 Epoch 30/30 6999/7000 ============================>. - ETA: 0s - loss: 0.0022 - accuracy: 0.9994 Epoch 00030: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0022 - accuracy: 0.9994 - val_loss: 0.0186 - val_accuracy: 0.9968 99.679 Epoch 1/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.1242 - accuracy: 0.9617 - val_loss: 0.0462 - val_accuracy: 0.9853 Epoch 2/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0532 - accuracy: 0.9841 - val_loss: 0.0348 - val_accuracy: 0.9899 Epoch 3/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0410 - accuracy: 0.9875 - val_loss: 0.0347 - val_accuracy: 0.9903 Epoch 4/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0343 - accuracy: 0.9894 - val_loss: 0.0277 - val_accuracy: 0.9922 Epoch 5/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0293 - accuracy: 0.9913 - val_loss: 0.0309 - val_accuracy: 0.9916 Epoch 6/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0266 - accuracy: 0.9922 - val_loss: 0.0262 - val_accuracy: 0.9923 Epoch 7/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0247 - accuracy: 0.9928 - val_loss: 0.0241 - val_accuracy: 0.9937 Epoch 8/30 7000/7000 ============================== - 22s 3ms/step - loss: 0.0222 - accuracy: 0.9934 - val_loss: 0.0250 - val_accuracy: 0.9932 Epoch 9/30 7000/7000 ============================== - 22s 3ms/step - loss: 0.0215 - accuracy: 0.9939 - val_loss: 0.0230 - val_accuracy: 0.9942 Epoch 10/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0206 - accuracy: 0.9942 - val_loss: 0.0249 - val_accuracy: 0.9933 Epoch 11/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0183 - accuracy: 0.9946 - val_loss: 0.0303 - val_accuracy: 0.9922 Epoch 12/30 6985/7000 ============================>. - ETA: 0s - loss: 0.0179 - accuracy: 0.9948 Epoch 00012: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0179 - accuracy: 0.9948 - val_loss: 0.0275 - val_accuracy: 0.9938 Epoch 13/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0102 - accuracy: 0.9970 - val_loss: 0.0183 - val_accuracy: 0.9958 Epoch 14/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0085 - accuracy: 0.9974 - val_loss: 0.0206 - val_accuracy: 0.9956 Epoch 15/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0084 - accuracy: 0.9976 - val_loss: 0.0196 - val_accuracy: 0.9958 Epoch 16/30 6987/7000 ============================>. - ETA: 0s - loss: 0.0075 - accuracy: 0.9977 Epoch 00016: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0075 - accuracy: 0.9977 - val_loss: 0.0207 - val_accuracy: 0.9957 Epoch 17/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0052 - accuracy: 0.9984 - val_loss: 0.0172 - val_accuracy: 0.9964 Epoch 18/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0045 - accuracy: 0.9986 - val_loss: 0.0153 - val_accuracy: 0.9966 Epoch 19/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0040 - accuracy: 0.9987 - val_loss: 0.0200 - val_accuracy: 0.9962 Epoch 20/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0041 - accuracy: 0.9988 - val_loss: 0.0194 - val_accuracy: 0.9965 Epoch 21/30 6998/7000 ============================>. - ETA: 0s - loss: 0.0034 - accuracy: 0.9990 Epoch 00021: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0034 - accuracy: 0.9990 - val_loss: 0.0188 - val_accuracy: 0.9964 Epoch 22/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0030 - accuracy: 0.9991 - val_loss: 0.0175 - val_accuracy: 0.9966 Epoch 23/30 7000/7000 ============================== - 23s 3ms/step - loss: 0.0029 - accuracy: 0.9991 - val_loss: 0.0185 - val_accuracy: 0.9966 Epoch 24/30 6986/7000 ============================>. - ETA: 0s - loss: 0.0024 - accuracy: 0.9992 Epoch 00024: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0024 - accuracy: 0.9992 - val_loss: 0.0164 - val_accuracy: 0.9969 Epoch 25/30 7000/7000 ============================== - 22s 3ms/step - loss: 0.0019 - accuracy: 0.9994 - val_loss: 0.0176 - val_accuracy: 0.9969 Epoch 26/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0020 - accuracy: 0.9995 - val_loss: 0.0189 - val_accuracy: 0.9968 Epoch 27/30 6997/7000 ============================>. - ETA: 0s - loss: 0.0019 - accuracy: 0.9993 Epoch 00027: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0019 - accuracy: 0.9993 - val_loss: 0.0173 - val_accuracy: 0.9970 Epoch 28/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0017 - accuracy: 0.9994 - val_loss: 0.0171 - val_accuracy: 0.9970 Epoch 29/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0018 - accuracy: 0.9995 - val_loss: 0.0165 - val_accuracy: 0.9972 Epoch 30/30 6983/7000 ============================>. - ETA: 0s - loss: 0.0016 - accuracy: 0.9995 Epoch 00030: ReduceLROnPlateau reducing learning rate to 1.5625000742147677e-05. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0016 - accuracy: 0.9995 - val_loss: 0.0172 - val_accuracy: 0.9971 99.707 Epoch 1/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.1187 - accuracy: 0.9635 - val_loss: 0.0407 - val_accuracy: 0.9872 Epoch 2/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0530 - accuracy: 0.9842 - val_loss: 0.0331 - val_accuracy: 0.9897 Epoch 3/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0405 - accuracy: 0.9877 - val_loss: 0.0309 - val_accuracy: 0.9906 Epoch 4/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0332 - accuracy: 0.9902 - val_loss: 0.0282 - val_accuracy: 0.9916 Epoch 5/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0308 - accuracy: 0.9910 - val_loss: 0.0221 - val_accuracy: 0.9936 Epoch 6/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0265 - accuracy: 0.9920 - val_loss: 0.0197 - val_accuracy: 0.9939 Epoch 7/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0249 - accuracy: 0.9925 - val_loss: 0.0198 - val_accuracy: 0.9946 Epoch 8/30 7000/7000 ============================== - 23s 3ms/step - loss: 0.0216 - accuracy: 0.9936 - val_loss: 0.0181 - val_accuracy: 0.9947 Epoch 9/30 7000/7000 ============================== - 22s 3ms/step - loss: 0.0203 - accuracy: 0.9941 - val_loss: 0.0221 - val_accuracy: 0.9942 Epoch 10/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0196 - accuracy: 0.9942 - val_loss: 0.0192 - val_accuracy: 0.9949 Epoch 11/30 6982/7000 ============================>. - ETA: 0s - loss: 0.0189 - accuracy: 0.9944 Epoch 00011: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0189 - accuracy: 0.9944 - val_loss: 0.0204 - val_accuracy: 0.9946 Epoch 12/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0106 - accuracy: 0.9968 - val_loss: 0.0148 - val_accuracy: 0.9962 Epoch 13/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0088 - accuracy: 0.9973 - val_loss: 0.0159 - val_accuracy: 0.9962 Epoch 14/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0085 - accuracy: 0.9975 - val_loss: 0.0163 - val_accuracy: 0.9962 Epoch 15/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0084 - accuracy: 0.9975 - val_loss: 0.0134 - val_accuracy: 0.9969 Epoch 16/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0075 - accuracy: 0.9979 - val_loss: 0.0165 - val_accuracy: 0.9965 Epoch 17/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0072 - accuracy: 0.9978 - val_loss: 0.0149 - val_accuracy: 0.9966 Epoch 18/30 6993/7000 ============================>. - ETA: 0s - loss: 0.0074 - accuracy: 0.9979 Epoch 00018: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0074 - accuracy: 0.9979 - val_loss: 0.0142 - val_accuracy: 0.9967 Epoch 19/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0044 - accuracy: 0.9986 - val_loss: 0.0140 - val_accuracy: 0.9970 Epoch 20/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0042 - accuracy: 0.9987 - val_loss: 0.0146 - val_accuracy: 0.9968 Epoch 21/30 6997/7000 ============================>. - ETA: 0s - loss: 0.0042 - accuracy: 0.9989 Epoch 00021: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0042 - accuracy: 0.9989 - val_loss: 0.0140 - val_accuracy: 0.9971 Epoch 22/30 7000/7000 ============================== - 23s 3ms/step - loss: 0.0030 - accuracy: 0.9991 - val_loss: 0.0134 - val_accuracy: 0.9971 Epoch 23/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0027 - accuracy: 0.9991 - val_loss: 0.0141 - val_accuracy: 0.9974 Epoch 24/30 6995/7000 ============================>. - ETA: 0s - loss: 0.0024 - accuracy: 0.9993 Epoch 00024: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05. 7000/7000 ============================== - 22s 3ms/step - loss: 0.0024 - accuracy: 0.9993 - val_loss: 0.0139 - val_accuracy: 0.9973 Epoch 25/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0025 - accuracy: 0.9993 - val_loss: 0.0136 - val_accuracy: 0.9973 Epoch 26/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0021 - accuracy: 0.9994 - val_loss: 0.0138 - val_accuracy: 0.9973 Epoch 27/30 6983/7000 ============================>. - ETA: 0s - loss: 0.0019 - accuracy: 0.9995 Epoch 00027: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0020 - accuracy: 0.9995 - val_loss: 0.0152 - val_accuracy: 0.9972 Epoch 28/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0018 - accuracy: 0.9995 - val_loss: 0.0142 - val_accuracy: 0.9974 Epoch 29/30 7000/7000 ============================== - 21s 3ms/step - loss: 0.0017 - accuracy: 0.9995 - val_loss: 0.0149 - val_accuracy: 0.9973 Epoch 30/30 6998/7000 ============================>. - ETA: 0s - loss: 0.0016 - accuracy: 0.9995 Epoch 00030: ReduceLROnPlateau reducing learning rate to 1.5625000742147677e-05. 7000/7000 ============================== - 21s 3ms/step - loss: 0.0016 - accuracy: 0.9995 - val_loss: 0.0141 - val_accuracy: 0.9973

1. Split the images

Predefined functions:

Tunning the digits width and visualizing the split process:

Finalizing the images:

Save the new datasets:

Validation:

Loading datasets:

Display the images in the new dataset:

2. Cleaning the images:

Import libraries

Loading datasets

Predefined functions:

Tunning the blur radiation:

Multi-object detection:

Save the new dataset:

Validation phase:

Load the dataset:

Display the images:

Building and optimizing a model

Import libraries

load the preprocessed datasets

building our model

Evaluate the model with 5-fold cross-validation

Ploting the learning curves of the model(s)

Setting up running

Saving the best model

Using our best model for predictions

Recommended on Deepnote

Stock Market Analysis

The 10 Best Ways to Create NumPy Arrays

Wide Residual Networks