import os, glob, time, random, argparse, cv2, numpy as np,PIL
from matplotlib import pyplot as plt
from sklearn.preprocessing import LabelEncoder
from sklearn.model_selection import train_test_splitpath="E:/python_projects/whale_dectection/trainimage"
# find .jpg and JPEG files in subdirectories
for subdir in os.listdir(path):
# find .jpg and JPEG files in subdirectories
for filename in glob.glob(os.path.join(path, subdir, '*.jpg')):
# read image
img = cv2.imread(filename)
# convert to RGB
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# pring their dimensions
#print(img.shape)
# reshape all images to a common size
img = cv2.resize(img, (224, 224))
# save image
cv2.imwrite(filename, img)
print(f" {filename} saved") E:/python_projects/whale_dectection/trainimage\beluga whale\beluga1.jpg saved
E:/python_projects/whale_dectection/trainimage\beluga whale\beluga2.jpg saved
E:/python_projects/whale_dectection/trainimage\beluga whale\beluga3.jpg saved
E:/python_projects/whale_dectection/trainimage\beluga whale\beluga4.jpg saved
E:/python_projects/whale_dectection/trainimage\beluga whale\beluga5.jpg saved
E:/python_projects/whale_dectection/trainimage\beluga whale\beluga6.jpg saved
E:/python_projects/whale_dectection/trainimage\beluga whale\beluga7.jpg saved
E:/python_projects/whale_dectection/trainimage\beluga whale\beluga8.jpg saved
E:/python_projects/whale_dectection/trainimage\beluga whale\beluga9.jpg saved
E:/python_projects/whale_dectection/trainimage\blue whale\blue1.JPG saved
E:/python_projects/whale_dectection/trainimage\blue whale\blue2.JPG saved
E:/python_projects/whale_dectection/trainimage\bowhead whale\bowhead1.jpg saved
E:/python_projects/whale_dectection/trainimage\fin whale\finn1.JPG saved
E:/python_projects/whale_dectection/trainimage\fin whale\finn2.JPG saved
E:/python_projects/whale_dectection/trainimage\ganges dolphin\ganges 1.jpg saved
E:/python_projects/whale_dectection/trainimage\ganges dolphin\ganges 2.jpg saved
E:/python_projects/whale_dectection/trainimage\ganges dolphin\ganges 3.jpg saved
E:/python_projects/whale_dectection/trainimage\ganges dolphin\ganges 4.jpg saved
E:/python_projects/whale_dectection/trainimage\ganges dolphin\ganges 5.jpg saved
E:/python_projects/whale_dectection/trainimage\ganges dolphin\ganges 6.jpg saved
E:/python_projects/whale_dectection/trainimage\gray whale\grey1.JPG saved
E:/python_projects/whale_dectection/trainimage\gray whale\grey2.JPG saved
E:/python_projects/whale_dectection/trainimage\humpback whale\humpback1.JPG saved
E:/python_projects/whale_dectection/trainimage\humpback whale\humpback2.JPG saved
E:/python_projects/whale_dectection/trainimage\humpback whale\humpback3.JPG saved
E:/python_projects/whale_dectection/trainimage\humpback whale\humpback4.JPG saved
E:/python_projects/whale_dectection/trainimage\humpback whale\humpback5.JPG saved
E:/python_projects/whale_dectection/trainimage\humpback whale\humpback6.JPG saved
E:/python_projects/whale_dectection/trainimage\humpback whale\humpback7.JPG saved
E:/python_projects/whale_dectection/trainimage\indoPacific dolphin\indopacific dolphin1.jpg saved
E:/python_projects/whale_dectection/trainimage\longfinnned whale\longfin1.JPG saved
E:/python_projects/whale_dectection/trainimage\river dolphin\river1.JPG saved
E:/python_projects/whale_dectection/trainimage\river dolphin\river10.JPG saved
E:/python_projects/whale_dectection/trainimage\river dolphin\river11.JPG saved
E:/python_projects/whale_dectection/trainimage\river dolphin\river12.JPG saved
E:/python_projects/whale_dectection/trainimage\river dolphin\river2.JPG saved
E:/python_projects/whale_dectection/trainimage\river dolphin\river3.JPG saved
E:/python_projects/whale_dectection/trainimage\river dolphin\river4.JPG saved
E:/python_projects/whale_dectection/trainimage\river dolphin\river5.JPG saved
E:/python_projects/whale_dectection/trainimage\river dolphin\river6.JPG saved
E:/python_projects/whale_dectection/trainimage\river dolphin\river7.JPG saved
E:/python_projects/whale_dectection/trainimage\river dolphin\river8.JPG saved
E:/python_projects/whale_dectection/trainimage\river dolphin\river9.JPG saved
E:/python_projects/whale_dectection/trainimage\sperm whale\sperm1.JPG saved
E:/python_projects/whale_dectection/trainimage\sperm whale\sperm2.JPG saved
E:/python_projects/whale_dectection/trainimage\sperm whale\sperm3.JPG saved
E:/python_projects/whale_dectection/trainimage\sperm whale\sperm4.JPG saved
E:/python_projects/whale_dectection/trainimage\sperm whale\sperm5.JPG saved
E:/python_projects/whale_dectection/trainimage\sperm whale\sperm6.JPG saved# make tensorflow model
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.applications.mobilenet_v2 import MobileNetV2
from tensorflow.keras.layers import Dropout, Flatten, Dense
from tensorflow.keras.models import Model
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.applications.mobilenet_v2 import preprocess_input
from tensorflow.keras.preprocessing.image import img_to_array
import numpy as np
import pandas as pd
import os
import tensorflow as tf
import keras
from keras.models import Sequential
from keras.layers import Dense, Dropout
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
# download VGG19 model
from keras.applications.vgg19 import VGG19# subdirectories as classelabel
classes = os.listdir(path)
species_to_label={}
for i, species in enumerate(classes):
species_to_label[species]=i
species_to_label
label_to_species={value:key for key,value in species_to_label.items()}
label_to_species{0: 'beluga whale',
1: 'blue whale',
2: 'bowhead whale',
3: 'fin whale',
4: 'ganges dolphin',
5: 'gray whale',
6: 'humpback whale',
7: 'indoPacific dolphin',
8: 'longfinnned whale',
9: 'river dolphin',
10: 'sperm whale'}df=[]
# make dataframe with path and label
for subdir in os.listdir(path):
for filename in glob.glob(os.path.join(path, subdir, '*.jpg')):
df.append([filename, subdir])
df=pd.DataFrame(df,columns=['path','label'])
df.head()| path | label | |
|---|---|---|
| 0 | E:/python_projects/whale_dectection/trainimage... | beluga whale |
| 1 | E:/python_projects/whale_dectection/trainimage... | beluga whale |
| 2 | E:/python_projects/whale_dectection/trainimage... | beluga whale |
| 3 | E:/python_projects/whale_dectection/trainimage... | beluga whale |
| 4 | E:/python_projects/whale_dectection/trainimage... | beluga whale |
def preprocess_image(path):
img=plt.imread(path)
im=cv2.resize(img,(224,224))
return im/255.0
# preprocess images
img=preprocess_image(df.loc[15,'path'])
label=df.loc[15,'label']
plt.imshow(img)
plt.title(label)Text(0.5, 1.0, 'ganges dolphin')
n_examples=len(df)
n_classes=len(classes)
# initialize X and Y
X=np.zeros((n_examples,224,224,3),dtype=np.float32)
Y=np.zeros((n_examples,n_classes),dtype=np.float32)
for i, (path,label) in enumerate(df.values):
X[i]=preprocess_image(path)
Y[i,species_to_label[label]]=1
print(X.shape)
print(Y.shape)(49, 224, 224, 3)
(49, 11)# split data into train and test
X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.2, random_state=42)# make model
model = Sequential()
base_model=VGG19(weights='imagenet',include_top=False,input_shape=(224,224,3),pooling='avg')
# add new layers
model.add(base_model)
model.add(Dense(1000, activation='relu', kernel_initializer='he_uniform'))
model.add(Dense(128, activation='relu', kernel_initializer='he_uniform'))
model.add(Dense(n_classes,activation='softmax'))
model.summary()Model: "sequential_6"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
vgg19 (Functional) (None, 512) 20024384
_________________________________________________________________
dense_17 (Dense) (None, 1000) 513000
_________________________________________________________________
dense_18 (Dense) (None, 128) 128128
_________________________________________________________________
dense_19 (Dense) (None, 11) 1419
=================================================================
Total params: 20,666,931
Trainable params: 20,666,931
Non-trainable params: 0
_________________________________________________________________batch_size = 16
epochs = 30
learning_rate = 1e-3
# compile model
model.compile(loss='categorical_crossentropy',optimizer='adam',metrics=['accuracy'])
# train model
history=model.fit(X_train,Y_train,epochs=epochs, batch_size=batch_size,validation_data=(X_test,Y_test))Epoch 1/30
3/3 [==============================] - 82s 24s/step - loss: 135.5160 - accuracy: 0.3077 - val_loss: 3.4185 - val_accuracy: 0.3000
Epoch 2/30
3/3 [==============================] - 67s 21s/step - loss: 5.0048 - accuracy: 0.0513 - val_loss: 2.5055 - val_accuracy: 0.0000e+00
Epoch 3/30
3/3 [==============================] - 80s 25s/step - loss: 2.1661 - accuracy: 0.2308 - val_loss: 2.8109 - val_accuracy: 0.1000
Epoch 4/30
3/3 [==============================] - 74s 22s/step - loss: 2.0730 - accuracy: 0.2821 - val_loss: 2.5080 - val_accuracy: 0.1000
Epoch 5/30
3/3 [==============================] - 76s 25s/step - loss: 2.0766 - accuracy: 0.3077 - val_loss: 2.6883 - val_accuracy: 0.1000
Epoch 6/30
3/3 [==============================] - 77s 23s/step - loss: 2.2028 - accuracy: 0.2308 - val_loss: 2.1229 - val_accuracy: 0.0000e+00
Epoch 7/30
3/3 [==============================] - 71s 22s/step - loss: 2.0935 - accuracy: 0.3333 - val_loss: 2.3295 - val_accuracy: 0.1000
Epoch 8/30
3/3 [==============================] - 81s 25s/step - loss: 2.0571 - accuracy: 0.2821 - val_loss: 2.5730 - val_accuracy: 0.1000
Epoch 9/30
3/3 [==============================] - 74s 22s/step - loss: 2.0983 - accuracy: 0.2821 - val_loss: 2.5278 - val_accuracy: 0.1000
Epoch 10/30
3/3 [==============================] - 63s 19s/step - loss: 2.0640 - accuracy: 0.2051 - val_loss: 2.3527 - val_accuracy: 0.0000e+00
Epoch 11/30
3/3 [==============================] - 70s 22s/step - loss: 2.0570 - accuracy: 0.2308 - val_loss: 2.4941 - val_accuracy: 0.1000
Epoch 12/30
3/3 [==============================] - 64s 19s/step - loss: 2.0568 - accuracy: 0.3333 - val_loss: 2.3143 - val_accuracy: 0.1000
Epoch 13/30
3/3 [==============================] - 62s 19s/step - loss: 2.1151 - accuracy: 0.2821 - val_loss: 2.2074 - val_accuracy: 0.1000
Epoch 14/30
3/3 [==============================] - 70s 22s/step - loss: 2.1120 - accuracy: 0.2821 - val_loss: 2.4766 - val_accuracy: 0.1000
Epoch 15/30
3/3 [==============================] - 64s 19s/step - loss: 2.0603 - accuracy: 0.2821 - val_loss: 2.5621 - val_accuracy: 0.1000
Epoch 16/30
3/3 [==============================] - 73s 22s/step - loss: 2.0687 - accuracy: 0.2308 - val_loss: 2.4641 - val_accuracy: 0.0000e+00
Epoch 17/30
3/3 [==============================] - 68s 21s/step - loss: 2.0662 - accuracy: 0.2308 - val_loss: 2.3878 - val_accuracy: 0.1000
Epoch 18/30
3/3 [==============================] - 77s 25s/step - loss: 2.0577 - accuracy: 0.2821 - val_loss: 2.4205 - val_accuracy: 0.1000
Epoch 19/30
3/3 [==============================] - 68s 19s/step - loss: 2.0544 - accuracy: 0.2308 - val_loss: 2.3655 - val_accuracy: 0.1000
Epoch 20/30
3/3 [==============================] - 70s 22s/step - loss: 2.0468 - accuracy: 0.3333 - val_loss: 2.3352 - val_accuracy: 0.1000
Epoch 21/30
3/3 [==============================] - 73s 23s/step - loss: 2.0125 - accuracy: 0.2821 - val_loss: 2.7639 - val_accuracy: 0.0000e+00
Epoch 22/30
3/3 [==============================] - 71s 23s/step - loss: 2.0100 - accuracy: 0.2564 - val_loss: 2.3822 - val_accuracy: 0.1000
Epoch 23/30
3/3 [==============================] - 93s 30s/step - loss: 2.0198 - accuracy: 0.3590 - val_loss: 3.3642 - val_accuracy: 0.0000e+00
Epoch 24/30
3/3 [==============================] - 81s 25s/step - loss: 1.9630 - accuracy: 0.3077 - val_loss: 3.6346 - val_accuracy: 0.1000
Epoch 25/30
3/3 [==============================] - 74s 23s/step - loss: 2.0163 - accuracy: 0.3590 - val_loss: 2.3716 - val_accuracy: 0.1000
Epoch 26/30
3/3 [==============================] - 84s 26s/step - loss: 2.3263 - accuracy: 0.2821 - val_loss: 2.1431 - val_accuracy: 0.0000e+00
Epoch 27/30
3/3 [==============================] - 76s 23s/step - loss: 2.1365 - accuracy: 0.1538 - val_loss: 2.4126 - val_accuracy: 0.1000
Epoch 28/30
3/3 [==============================] - 82s 26s/step - loss: 2.1137 - accuracy: 0.2821 - val_loss: 2.4048 - val_accuracy: 0.1000
Epoch 29/30
3/3 [==============================] - 76s 22s/step - loss: 2.0445 - accuracy: 0.2564 - val_loss: 2.3340 - val_accuracy: 0.0000e+00
Epoch 30/30
3/3 [==============================] - 77s 25s/step - loss: 2.0986 - accuracy: 0.2308 - val_loss: 2.3615 - val_accuracy: 0.0000e+00# plot loss and accuracy 2,2 subplot
plt.subplot(2,2,1)
plt.plot(history.history['loss'],label='loss')
plt.plot(history.history['val_loss'],label='val_loss')
plt.xlabel('epoch')
plt.legend()
plt.subplot(2,2,2)
plt.plot(history.history['accuracy'],label='accuracy')
plt.plot(history.history['val_accuracy'],label='val_accuracy')
plt.legend()<matplotlib.legend.Legend at 0x2074c092130>
# save model
model.save('model.h5')
# load model
from tensorflow.keras.models import load_model
model=load_model('model.h5')# make predictions
plt.imshow(X_test[1])
plt.title(label_to_species[np.argmax(Y_test[1])]+'_model_prediction_'+label_to_species[np.argmax(model.predict(X_test[1:2]))])
# print test accuracy two decimal places with fsprintf
#print('Test accuracy: %.2f' % (model.evaluate(X_test,Y_test)[1]*100))
# print test accuracy two decimal places with f string
print(f'Test accuracy: {model.evaluate(X_test,Y_test)[1]*100:.3f}')1/1 [==============================] - 3s 3s/step - loss: 2.3615 - accuracy: 0.0000e+00
Test accuracy: 0.000
def model_predict(path):
img=plt.imread(path)
img=cv2.resize(img,(224,224))
img=img/255.0
img=img.reshape(1,224,224,3)
return label_to_species[np.argmax(model.predict(img))]
model_predict('E:/python_projects/whale_dectection/newimage/ganges 5.jpg')'beluga whale'def model_predict(path):
im = np.array([preprocess_image(path)])
prediction = model.predict(im)
# return prediction label and probability
return label_to_species[np.argmax(prediction)], np.max(prediction)
plt.figure(figsize=(10, 10))
for n, path in enumerate(os.listdir('E:/python_projects/whale_dectection/newimage')):
prediction, confidence = model_predict('E:/python_projects/whale_dectection/newimage/' + path)
plt.subplot(2, 2, n+1)
# add title to each subplot
plt.title(f'model predicts {prediction} ({confidence*100:.2f}%)')
plt.imshow(preprocess_image('E:/python_projects/whale_dectection/newimage/' + path))