def loadled():
path = "/Users/wzq/Desktop/led/train"
#dirs=os.listdir(path)
label = []
image = []
for i in range(1, 10):
dir_image = os.listdir(path + "/" + str(i))
for j in dir_image:
if j in [".DS_Store"]:
continue
gray = cv2.imread(path + "/" + str(i) + "/" + j,
cv2.IMREAD_GRAYSCALE)
bw = cv2.adaptiveThreshold(gray, 255,
cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY, 25, 25)
bw = resize(bw)
# cv2.imshow("",bw)
# cv2.waitKey(0)
# gray = cv2.resize(img, (28, 28), interpolation=cv2.INTER_AREA)
# gray = gray[:, :, 0]
# fd = hog(bw.reshape((28, 28)), orientations=9, pixels_per_cell=(1, 1), cells_per_block=(1, 1),
# visualise=False)
label.append(i)
image.append(bw)
return image, label
def create_training_data():
image = []
for person in Person:
path = os.path.join(file, person)
class_num = Person.index(person)
for img in os.listdir(path):
img_array = cv2.imread(os.path.join(path, img))
new_array = cv2.resize(img_array, (224, 224))
image.append([new_array, class_num])
return image
data = pd.read_csv('mapping.csv') # reading the csv file
data.head() # printing first five rows of the file
X = [ ] # creating an empty array
for img_name in data.Image_ID:
img = plt.imread('' + img_name)
X.append(img) # storing each image in array X
X = np.array(X) # converting list to array
y = data.Class
dummy_y = np_utils.to_categorical(y) # one hot encoding Classes
image = []
for i in range(0,X.shape[0]):
a = resize(X[i], preserve_range=True, output_shape=(224,224)).astype(int) # reshaping to 224*224*3
image.append(a)
X = np.array(image)
from keras.applications.vgg16 import preprocess_input
X = preprocess_input(X, mode='tf') # preprocessing the input data
from sklearn.model_selection import train_test_split
X_train, X_valid, y_train, y_valid = train_test_split(X, dummy_y, test_size=0.3, random_state=42) # preparing the validation set
from keras.models import Sequential
from keras.applications.vgg16 import VGG16
from keras.layers import Dense, InputLayer, Dropout
https://www.kaggle.com/rhammell/ships-in-satellite-imagery
* initial attempt to classify based on the given lables (partial ship images classfied as '0')
* I made the dataset more complicated by reclassifying partial ships as '1' label; note that ship images crossing the corner
were typically classified as '0' unless it's very clear the image is a ship (typically needs to include a bow or a stern)
* poor initial results addressed by building more data for misclassified types of images
* increased size of model to address the increased image complexity
* 91%+ prediction on the test set and can be further improved
'''
np.random.seed(88)
#upload RGB images
image = []
for image_path in glob.glob("/Users/vlad/Projects/ships-in-satellite-imagery/shipsnet/*.png"):
image.append(misc.imread(image_path))
X_orig = np.asarray(image)
#get list of file names
filename = []
filename.append(glob.glob("/Users/vlad/Projects/ships-in-satellite-imagery/shipsnet/*.png"))
filename = np.asarray(filename).T
#60th position in line corresponds to the label value => create labels list
y_orig = []
for n in range(len(filename)):
line = np.array_str(filename[n])
y_orig.append(line[60])
y_orig = np.asarray(y_orig).T
y_orig = y_orig.reshape((len(y_orig), 1))
def my_input(url):
label_binarizer_classes_ = [
'Pepper__bell___Bacterial_spot', 'Pepper__bell___healthy',
'Potato___Early_blight', 'Potato___Late_blight', 'Potato___healthy',
'Tomato_Bacterial_spot', 'Tomato_Early_blight', 'Tomato_Late_blight',
'Tomato_Leaf_Mold', 'Tomato_Septoria_leaf_spot',
'Tomato_Spider_mites_Two_spotted_spider_mite', 'Tomato__Target_Spot',
'Tomato__Tomato_YellowLeaf__Curl_Virus', 'Tomato__Tomato_mosaic_virus',
'Tomato_healthy'
]
model = Sequential()
inputShape = (256, 256, 3)
chanDim = -1
if K.image_data_format() == "channels_first":
inputShape = (3, 256, 256)
chanDim = 1
model.add(Conv2D(32, (3, 3), padding="same", input_shape=inputShape))
model.add(Activation("relu"))
model.add(BatchNormalization(axis=chanDim))
model.add(MaxPooling2D(pool_size=(3, 3)))
model.add(Dropout(0.25))
model.add(Conv2D(64, (3, 3), padding="same"))
model.add(Activation("relu"))
model.add(BatchNormalization(axis=chanDim))
model.add(Conv2D(64, (3, 3), padding="same"))
model.add(Activation("relu"))
model.add(BatchNormalization(axis=chanDim))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Conv2D(128, (3, 3), padding="same"))
model.add(Activation("relu"))
model.add(BatchNormalization(axis=chanDim))
model.add(Conv2D(128, (3, 3), padding="same"))
model.add(Activation("relu"))
model.add(BatchNormalization(axis=chanDim))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(1024))
model.add(Activation("relu"))
model.add(BatchNormalization())
model.add(Dropout(0.5))
model.add(Dense(15))
model.add(Activation("softmax"))
opt = Adam(lr=1e-3, decay=1e-3 / 25)
model.compile(loss="binary_crossentropy",
optimizer=opt,
metrics=["accuracy"])
model.load_weights(
"C:/Users/Rajan Sethi/Desktop/Online Courses/mangobite/mangobite/weights_test_cnn.hdfs"
)
image = list()
test = convert_image_to_array(url)
image.append(test)
np_image_test = np.array(image, dtype=np.float16) / 225.0
predictions_1 = model.predict(np_image_test)
predictions = np.argmax(predictions_1, 1)
predictions[0]
tf.keras.backend.clear_session()
K.clear_session()
return label_binarizer_classes_[predictions[0]]
