def load_img_scaled(input_path, target_shape, grayscale=False):
return np.expand_dims(
image.img_to_array(
image.load_img(
input_path, target_size=target_shape, grayscale=grayscale)) /
255.0,
axis=0)
def load_img(input_path, target_shape, grayscale=False, mean=None, std=None):
img = image.load_img(
input_path, target_size=target_shape, grayscale=grayscale)
img_arr = np.expand_dims(image.img_to_array(img), axis=0)
if not grayscale:
img_arr = preprocess_input(img_arr, mean=mean, std=std)
return img_arr
if args["image"]:
img = cv2.imread(args["image"])
height, width = img.shape[:2]
if width > 750 or width < 150:
img = imutils.resize(img, width=600)
height, width = img.shape[:2]
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
img_clone = img.copy()
rects = face_cascade.detectMultiScale(gray,
scaleFactor=1.1,
minNeighbors=5)
if len(rects) == 0:
roi = cv2.resize(gray, (28, 28))
roi = roi.astype("float") / 255.0
roi = img_to_array(roi)
roi = np.expand_dims(roi, axis=0)
notSmiling, smiling = model.predict(roi)[0]
label = "Smiling" if smiling > notSmiling else "Not Smiling"
cv2.putText(img_clone, f"{label}: {max(notSmiling, smiling) * 100}%",
(width // 4, height // 2), cv2.FONT_HERSHEY_SIMPLEX, 0.7,
(255, 0, 25), 2)
for fX, fY, fW, fH in rects:
roi = gray[fY:fY + fH, fX:fX + fW]
roi = cv2.resize(roi, (28, 28))
roi = roi.astype("float") / 255.0
roi = img_to_array(roi)
roi = np.expand_dims(roi, axis=0)
notSmiling, smiling = model.predict(roi)[0]
label = "Smiling" if smiling > notSmiling else "Not Smiling"
textpos = (fX, fY - 10)
print('new_learning:', init_lr)
if num_iter == int(2*train_iter/3):
init_lr = init_lr/10
model.compile(optimizer=Adam(learning_rate=init_lr), loss='mse')
print('new_learning:', init_lr)
#randomly select a batch of sample
select_case = [np.random.randint(1,len(all_label)) for _ in range(batch_size)]
batch_input = []
batch_output = []
for i in range(batch_size):
all_sensor_input = np.zeros((num_sensors, 84, 84*4, 3))
all_sensor_output = np.zeros((num_sensors, 1))
for idx_sensor in range(num_sensors):
sensor_path = 'training_random/' + all_sensors[idx_sensor]
img_1 = image.load_img(sensor_path+'/1/'+filelist[select_case[i]], target_size=(84,84)) #height-width
img_array_1 = image.img_to_array(img_1)
img_2 = image.load_img(sensor_path+'/2/'+filelist[select_case[i]], target_size=(84,84)) #height-width
img_array_2 = image.img_to_array(img_2)
img_3 = image.load_img(sensor_path+'/3/'+filelist[select_case[i]], target_size=(84,84)) #height-width
img_array_3 = image.img_to_array(img_3)
img_4 = image.load_img(sensor_path+'/4/'+filelist[select_case[i]], target_size=(84,84)) #height-width
img_array_4 = image.img_to_array(img_4)
all_sensor_input[idx_sensor,:, 84*3:84*4,:] = img_array_4/255
all_sensor_input[idx_sensor,:, 84*2:84*3,:] = img_array_3/255
all_sensor_input[idx_sensor,:, 84*1:84*2,:] = img_array_2/255
all_sensor_input[idx_sensor,:, 84*0:84*1,:] = img_array_1/255
batch_input.append(all_sensor_input.copy())
# get label data
img_index = int(filelist[select_case[i]][:-4])
batch_output.append(all_label[select_case[i]])
batch_input = np.array(batch_input)
from os import listdir, remove
from numpy import asarray
from numpy import save
from tensorflow.keras.preprocessing.image import load_img
from tensorflow.keras.preprocessing.image import img_to_array
from PIL import Image
import numpy as np
for x in listdir("../q7_img"):
for z in listdir("../q7_img/" + x):
ind = 0
for y in listdir("../q7_img/" + x + '/' + z):
ind += 1
p = "../q7_img/" + x + '/' + z + '/' + y
print(p)
arr = load_img(p, target_size=(224, 224))
arr = img_to_array(arr).astype(np.uint8)
print(arr)
im = Image.fromarray(arr)
remove(p)
im.save("../q7_img/" + x + '/' + z + '/' + str(ind) + ".jpg")
print("\nModel:", model_name)
print("Model version:", model_version)
print("Image:", image_path)
print("Port:", port)
host = "127.0.0.1"
port = port
server = host + ':' + port
model_name = model_name
model_version = int(model_version)
request_timeout = float(10)
image_filepaths = [image_path]
# Loading image
test_image = image.load_img(image_filepaths[0], target_size=(224, 224))
test_image = image.img_to_array(test_image)
test_image = test_image.astype('float32')
test_image = test_image / 255.0
# Create gRPC client and request
channel = grpc.insecure_channel(server)
stub = prediction_service_pb2_grpc.PredictionServiceStub(channel)
request = predict_pb2.PredictRequest()
request.model_spec.name = model_name
request.model_spec.version.value = model_version
request.model_spec.signature_name = "serving_default"
request.inputs['vgg16_input'].CopyFrom(
tensor_util.make_tensor_proto(test_image,
shape=[1] + list(test_image.shape)))
def splitSearch(self, markerImg):
rightNow = rospy.Time.now()
IMG_HEIGHT, IMG_WIDTH, colors = markerImg.shape
tiles = []
results = []
has_marker_chance = [] # for debugging
no_marker_percent = [] #for debuging
# split into 9 tiles
step_width = IMG_WIDTH // 3
step_height = IMG_HEIGHT // 3
if (VERBOSE):
now = str(rospy.Time.now)
cv2.imshow("on Laptop ", markerImg)
cv2.waitKey(1)
if (SPAM):
rospy.loginfo("IMG: {}x{}".format(round(IMG_WIDTH, 2),
round(IMG_HEIGHT, 2)))
for i in range(3):
for j in range(3):
x_start, x_end = i * step_width, (i + 1) * step_width
y_start, y_end = j * step_height, (j + 1) * step_height
tiles.append(
cv2.resize(markerImg[x_start:x_end, y_start:y_end],
(step_width, step_height)))
'''
if(VERBOSE):
rospy.loginfo(len(tiles))
for x in range(9):
cv2.imshow(str(x),tiles[x] )
cv2.waitKey(0)
#cv2.destroyAllWindows()
'''
if (SPAM):
rospy.loginfo("{}:{}x{}:{}".format(round(x_start, 2),
round(x_end, 2),
round(y_start, 2),
round(y_end, 2)))
certaintyLevel = 5
#threadsavety in Python is a nightmare
with graph.as_default():
with thread_session.as_default():
for tile in tiles:
tile = numpy.expand_dims(img_to_array(
tile.astype("float") / 255.0),
axis=0)
#cv2.imshow('',tile)
#cv2.waitKey(0)
(noMarker, marker) = model.predict(tile)[0]
probability = round((max(noMarker, marker) * 100), 2)
certainty = ((marker - noMarker) * 100)
if (SPAM): #output the difference in certainty
rospy.loginfo("degree of certainty")
rospy.loginfo(certainty)
has_marker_chance.append(marker * 100) #fill debug array
no_marker_percent.append(noMarker * 100) #fill debug array
# do not trust if marker > no marker
if ((marker > noMarker)
and (certainty > certaintyLevel)): # bad bad BUG!
has_marker = True
else:
has_marker = False
results.append([has_marker, probability])
#output formated array content of AI results
if (VERBOSE):
rospy.loginfo("certainty of Marker found")
rospy.loginfo("{:6.2f} {:6.2f} {:6.2f}".format(
has_marker_chance[0], has_marker_chance[3],
has_marker_chance[6]))
rospy.loginfo("{:6.2f} {:6.2f} {:6.2f}".format(
has_marker_chance[1], has_marker_chance[4],
has_marker_chance[7]))
rospy.loginfo("{:6.2f} {:6.2f} {:6.2f}".format(
has_marker_chance[2], has_marker_chance[5],
has_marker_chance[8]))
rospy.loginfo(" percent certainty that there is No Marker")
rospy.loginfo("{:6.2f} {:6.2f} {:6.2f}".format(
no_marker_percent[0], no_marker_percent[3],
no_marker_percent[6]))
rospy.loginfo("{:6.2f} {:6.2f} {:6.2f}".format(
no_marker_percent[1], no_marker_percent[4],
no_marker_percent[7]))
rospy.loginfo("{:6.2f} {:6.2f} {:6.2f}".format(
no_marker_percent[2], no_marker_percent[5],
no_marker_percent[8]))
rightDuration = rospy.Time.now() - rightNow
rospy.logwarn(
"AI took {} Sec ({} nSec) to process the image".format(
rightDuration.to_sec(), rightDuration.to_nsec()))
return results, has_marker_chance, no_marker_percent
interpreter1.allocate_tensors()
input_details1 = interpreter1.get_input_details()
output_details1 = interpreter1.get_output_details()
#This is interpreter for GRU_8_8
interpreter2 = tf.lite.Interpreter(model_path='gru.tflite')
interpreter2.allocate_tensors()
input_details2 = interpreter2.get_input_details()
output_details2 = interpreter2.get_output_details()
feature_seq = list() #This will containg[1,20,7*7*1280]
for i in range(1, 21):
seq = list() #List of frame
img = image.load_img('pic/{}.jpg'.format(i), target_size=(224, 224))
img = image.img_to_array(img, dtype=np.float32)
seq.append(img)
seq = np.array(seq, dtype=np.float32)
interpreter1.set_tensor(input_details1[0]["index"],
seq) # seq is the input
interpreter1.invoke()
feature = interpreter1.get_tensor(output_details1[0]["index"])
print("feature {} extracted".format(i))
feature = feature.reshape(-1)
feature_seq.append(feature)
feature_seq = np.array(feature_seq, dtype=np.float32)
feature_seq = feature_seq.reshape(-1, 20, 7 * 7 * 1280)
print(feature_seq.shape)
def detect_and_predict_mask(frame, faceNet, maskNet):
# lay kich thuoc cua frame va khoi tao blob
# cho frame
(h, w) = frame.shape[:2]
blob = cv2.dnn.blobFromImage(frame, 1.0, (300, 300), (104.0, 177.0, 123.0))
# Cho blob vao mo hinh phat hien khuon mat
faceNet.setInput(blob)
detections = faceNet.forward()
# Khoi tao lists chua anh khuon mat, vi tri tuong ung cua khuon mat,
# danh sach cac du doan khuon mat
faces = []
locs = []
preds = []
name = ""
# lap qua cac phat hien khuon mat
for i in range(0, detections.shape[2]):
# trich xuat do tin cay (i.e., phan tram) cua phat hien
# khuon mat
confidence = detections[0, 0, i, 2]
# loc cac phat hien yeu bang cach dam bao do tin cay lon hon
# do tin cay toi thieu
if confidence > args["confidence"]:
# tinh toa do (x, y)-cua hop gioi han cho doi tuong
box = detections[0, 0, i, 3:7] * np.array([w, h, w, h])
(startX, startY, endX, endY) = box.astype("int")
# dam bao cac hop gioi han nam trong kich thuoc cua
# frame anh
(startX, startY) = (max(0, startX), max(0, startY))
(endX, endY) = (min(w - 1, endX), min(h - 1, endY))
# Trich xuat ROI cua khuon mat, chuyen doi tu he mau BGR qua
# he mau RGB, thay doi kich thuoc ve 224 x 224 va xu ly truoc
face = frame[startY:endY, startX:endX]
# ensure the face width and height are sufficiently large
# if fW < 20 or fH < 20:
# continue
# construct a blob for the face ROI, then pass the blob
# through our face embedding model to obtain the 128-d
# quantification of the face
faceBlob = cv2.dnn.blobFromImage(face,
1.0 / 255, (96, 96), (0, 0, 0),
swapRB=True,
crop=False)
embedder.setInput(faceBlob)
vec = embedder.forward()
# perform classification to recognize the face
preds = recognizer.predict_proba(vec)[0]
j = np.argmax(preds)
proba = preds[j]
name = le.classes_[j]
# chuyen doi tu he mau BGR qua he mau RGB, thay doi kich thuoc
# ve 224 x 224 va xu ly truoc
face = cv2.cvtColor(face, cv2.COLOR_BGR2RGB)
face = cv2.resize(face, (224, 224))
face = img_to_array(face)
face = preprocess_input(face)
# Them khuon mat vao faces va toa do hop gioi han vao locs
faces.append(face)
locs.append((startX, startY, endX, endY))
# Chi dua ra du doan neu mot khuon mat duoc phat hien
if len(faces) > 0:
# dua ra du doan ve tat ca *all*
# khuon mat cung mot luc thay ve du doan tung khuong mat
# trong vong lap `for` o tren
faces = np.array(faces, dtype="float32")
preds = maskNet.predict(faces, batch_size=32)
# tra ve 2 bo vi tri khuon mat va vi tri tuong ung
# cua chung
return (locs, preds, name)
def evaluate_model(list_number=1):
print("👾 evaluating model on list number {}".format(list_number))
print("👾 loading weights")
model = tf.compat.v1.keras.experimental.load_from_saved_model(
path.join(MODEL_ROOT, "model"))
model.build(IMAGE_SIZE + (3, ))
print(model.summary())
print("👾 reading test file")
file = open(
path.join(ANNOTATIONS_ROOT, "testlist0{}.txt".format(list_number)),
"r")
temp = file.read()
videos = temp.split('\n')
# pick up only the classes I need
videos = list(
set([a for a in videos for b in CLASSES if b in a.split("/")[0]]))
# creating the dataframe
print("👾 creating dataframe from test file")
test = pd.DataFrame()
test['video'] = videos
test = test[:-1]
test_videos = test['video']
# creating the tags
print("👾 generating tags for labels")
train = pd.read_csv(path.join(ANNOTATIONS_ROOT, 'trainlist01_frames.csv'))
y = train['label']
y = pd.get_dummies(y)
# creating two lists to store predicted and actual tags
predict = []
actual = []
# checking if temp root exist otherwise create it
if not path.exists("temp"):
print("👾 creating folder temp")
os.makedirs("temp")
# for loop to extract frames from each test video
print("👾 extracting frames from test videos")
for i in tqdm(range(test_videos.shape[0])):
count = 0
videoFile = test_videos[i]
cap = cv2.VideoCapture(path.join(
VIDEOS_ROOT, videoFile)) # capturing the video from the given path
frameRate = cap.get(5) #frame rate
x = 1
# removing all other files from the temp folder
files = glob('temp/*')
for f in files:
os.remove(f)
while (cap.isOpened()):
frameId = cap.get(1) #current frame number
ret, frame = cap.read()
if (ret != True):
break
if (frameId % math.floor(frameRate) == 0):
# storing the frames of this particular video in temp folder
filename = 'temp/' + "_frame%d.jpg" % count
count += 1
cv2.imwrite(filename, frame)
cap.release()
# reading all the frames from temp folder
images = glob("temp/*.jpg")
prediction_images = []
for i in range(len(images)):
img = load_img(images[i], target_size=IMAGE_SIZE + (3, ))
img = img_to_array(img)
prediction_images.append(img)
# converting all the frames for a test video into numpy array
prediction_images = np.array(prediction_images)
# predicting tags for each array
prediction = np.argmax(model.predict(prediction_images), axis=-1)
print(prediction)
# appending the mode of predictions in predict list to assign the tag to the video
predict.append(y.columns.values[s.mode(prediction)[0][0]])
# appending the actual tag of the video
actual.append(videoFile.split('/')[0])
print("evaluation done")
print(accuracy_score(predict, actual) * 100)
endX = (x + 1) * stepX
endY = (y + 1) * stepY
# add the (x, y)-coordinates to our cell locations list
row.append((startX, startY, endX, endY))
# crop the cell from the warped transform image and then
# extract the digit from the cell
cell = warped[startY:endY, startX:endX]
digit = get_digit(cell)
if digit is not None:
digit_preprocessed = cv2.resize(digit, (28, 28))
digit_preprocessed = digit_preprocessed.astype("float") / 255.0
digit_preprocessed = img_to_array(digit_preprocessed)
digit_preprocessed = np.expand_dims(digit_preprocessed, axis=0)
pred = model.predict(digit_preprocessed).argmax()
board[y, x] = pred
# add the row to our cell locations
cellLocs.append(row)
for i in board:
for y in i:
if y == 0:
print(y)
solve(board)
def preprocessing_func(image):
'''Function that preprocesses the input'''
preproc_img = img_to_array(image)
preproc_img = np.expand_dims(preproc_img, axis = 0)
preproc_img = preprocess_input(preproc_img)
return preproc_img
def load_data(directory,
classes,
rescale=True,
preprocess=None,
verbose=False):
""" Helper function to load data in a Keras friendly format """
if not os.path.exists(directory):
raise FileNotFoundError(directory + ' not found')
# Check directories
directories = os.listdir(directory)
dataset = list()
for d in directories:
if d not in classes or not os.path.isdir(os.path.join(directory, d)):
print('Skipping', d)
continue
if verbose:
print('Loading directory', d)
for f in os.listdir(os.path.join(directory, d)):
try:
# Load image
img = load_img(os.path.join(os.path.join(directory, d), f),
color_mode='rgb',
target_size=[256, 256])
except PIL.UnidentifiedImageError:
continue
# Convert to numpy array
img = img_to_array(img)
# Apply any preprocess function and rescaling
if preprocess is not None:
img = preprocess(img)
else:
if rescale:
img /= 255
# Get index number
class_num = classes.index(d)
# Append
dataset.append([img, class_num])
# Shuffle dataset
random.shuffle(dataset)
x = list()
y = list()
# Create our x and y arrays
for img, label in dataset:
x.append(img)
y.append(label)
# Reshape x,y into the required formats
x = np.array(x).reshape(-1, 256, 256, 3)
y = to_categorical(y)
return x, y, 0
def image_to_tensor(image_path, image_size):
img = image.load_img(image_path, target_size=image_size[:2])
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
return x
classes=['Brabo','Feliz','Neutro']
while(True):
ret, img = cap.read()
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray, 1.3, 5)
#print(faces)
for (x,y,w,h) in faces:
cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)
rosto = img[y:y+h,x:x+w]
rosto = cv2.resize(rosto, (100, 100))
rosto = image.img_to_array(rosto)
rosto = rosto.reshape((1,) + rosto.shape)
predict = model.predict_classes(rosto)
print(classes[predict[0]])
cv2.putText(img, classes[predict[0]], (x, y), cv2.FONT_HERSHEY_SIMPLEX, 1.2, (0,255,0), 2)
print('Círculo')
cv2.imshow('Image',img)
if cv2.waitKey(1) == 27:
break
cap.release()
cv2.destroyWindow('Image')
cv2.destroyAllWindows()
def load_image(img_path):
img = image.load_img(img_path, target_size=(224, 224))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = imagenet_utils.preprocess_input(x)
return x
use_batch_norm=False,
num_classes=1,
filters=64,
dropout=0.2,
output_activation='sigmoid')
model.compile(optimizer='Adam', loss='MeanSquaredError')
print(np.shape(x_train), " ", np.shape(y_train))
model.fit(x=x_train, y=y_train, batch_size=32, epochs=5, validation_split=.2)
#input_img = image.load_img('./love-me-do.png', target_size=(640,480))
input_img = pixels_from_path('./love-me-do.jpg')
print("input shape", np.array(input_img).shape)
img_array = image.img_to_array(input_img)
print("input array", np.array(img_array).shape)
img_batch = np.expand_dims(img_array, axis=0)
print("input batch", np.array(img_batch).shape)
img_preprocessed = preprocess_input(img_batch)
print("input pre", np.array(img_preprocessed).shape)
predictions = model.predict(img_preprocessed)
print("predictions shape is ", predictions.shape)
print("prediction is ", predictions)
test = np.reshape(predictions, (1120, 1280))
print("Test ", test.shape)
print(test)
if player_num == 1:
while globalvar.leave_game != True:
# 從WebCam讀取一張圖片
suc, frame_A = camera_A.read()
# 顯示圖片
cv2.imshow('frame', frame_A)
cv2.waitKey(100)
frame_A = cv2.cvtColor(frame_A, cv2.COLOR_BGR2RGB)
frame_A = cv2.resize(frame_A, (img_x, img_y), interpolation=cv2.INTER_NEAREST)
frame_A = np.array(frame_A) / 255
frame_A = image.img_to_array(frame_A)
frame_A = np.expand_dims(frame_A, axis = 0)
predict_output_A = model.predict(frame_A)
max_probablility_index_A = predict_output_A.argmax()
max_probablility_A = np.max(predict_output_A)
print("Player A:", label_list[max_probablility_index_A], ":", "{:.2%}".format(max_probablility_A), end="\r")
if player_num == 2:
while globalvar.leave_game != True:
# 從WebCam讀取一張圖片
suc, frame_A = camera_A.read()
suc, frame_B = camera_B.read()
else:
ret, img = cam.read()
if not ret:
print('error')
break
key = cv2.waitKey(1)
if key == 27: # when ESC key is pressed break
break
count += 1
if count > max_count:
X = []
img_org = cv2.resize(img, (640, 480))
img = cv2.resize(img, (256, 256))
img = img_to_array(img)
X.append(img)
X = np.asarray(X)
X = X / 255.0
start = time.time()
preds = model_pred.predict(X)
elapsed_time = time.time() - start
pred_label = ''
label_num = 0
tmp_max_pred = 0
print(preds)
for i in preds[0]:
if i > tmp_max_pred:
pred_label = labels[label_num]
def process_image(img_path):
img = image.load_img(img_path, target_size=(224, 224))
img_array = image.img_to_array(img)
img_array = np.expand_dims(img_array, axis=0)
pImg = mobilenet_v2.preprocess_input(img_array)
return pImg
box = detections[0, 0, i, 3:7] * np.array([w, h, w, h])
(startX, startY, endX, endY) = box.astype("int")
# ensure the detected bounding box does fall outside the
# dimensions of the frame
startX = max(0, startX)
startY = max(0, startY)
endX = min(w, endX)
endY = min(h, endY)
# extract the face ROI and then preproces it in the exact
# same manner as our training data
face = frame[startY:endY, startX:endX]
face = cv2.resize(face, (32, 32))
face = face.astype("float") / 255.0
face = img_to_array(face)
face = np.expand_dims(face, axis=0)
# pass the face ROI through the trained liveness detector
# model to determine if the face is "real" or "fake"
preds = model.predict(face)[0]
j = np.argmax(preds)
label = le.classes_[j]
if label == 'fake':
label = "{}: {:.4f}".format(label, preds[j])
cv2.putText(frame, label, (startX, startY - 10),
cv2.FONT_HERSHEY_SIMPLEX, 5, (0, 0, 255), 8)
cv2.rectangle(frame, (startX, startY), (endX, endY),
(0, 0, 255), 2)
else:
labels = []
# with open(backup_dir + '/labels.txt','r') as f:
with open(args.labels, 'r') as f:
for line in f:
labels.append(line.rstrip())
print(labels)
model_pred = model_from_json(open(args.model).read())
model_pred.load_weights(args.weights)
# model_pred.summary()
X = []
img_path = args.testfile
img = img_to_array(load_img(img_path, target_size=(64, 64)))
X.append(img)
X = np.asarray(X)
preds = model_pred.predict(X)
pred_label = ""
label_num = 0
for i in preds[0]:
if i == 1.0:
pred_label = labels[label_num]
break
label_num += 1
print("label=" + pred_label)
start_time = time.time()
#load image
img_path = "cat.jpg" #animal class
#img_path = "other.jpg" #other class
#img_path = "person.jpg" #person class
#resize image
img = load_img(img_path, target_size=(299, 299))
#show image
plt.imshow(img)
plt.show()
#image to array
new_img = image.img_to_array(img)
new_img /= 255
new_img = np.expand_dims(new_img, axis=0)
# input_details[0]['index'] = the index which accepts the input
interpreter.set_tensor(input_details[0]['index'], new_img)
# run the inference
interpreter.invoke()
# The function `get_tensor()` returns a copy of the tensor data.
# Use `tensor()` in order to get a pointer to the tensor.
output_data = interpreter.get_tensor(output_details[0]['index'])
#print(output_data)
#stop time
def prepare_image(file):
img_path = 'E:/Stay Away You!/DL_training/Mobile_Net/data/Mobilenet_sample/'
img = image.load_img(img_path + file,target_size=(224,224))
img_array = image.img_to_array(img)
img_array_expand_dimension = np.expand_dims(img_array,axis = 0)
return tf.keras.applications.mobilenet.preprocess_input(img_array_expand_dimension)
import random from tensorflow.keras.preprocessing.image import img_to_array, load_img # Let's define a new Model that will take an image as input, and will output # intermediate representations for all layers in the previous model after # the first. successive_outputs = [layer.output for layer in model.layers[1:]] visualization_model = Model(img_input, successive_outputs) # Let's prepare a random input image of a cat or dog from the training set. cat_img_files = [os.path.join(train_cats_dir, f) for f in train_cat_fnames] dog_img_files = [os.path.join(train_dogs_dir, f) for f in train_dog_fnames] img_path = random.choice(cat_img_files + dog_img_files) img = load_img(img_path, target_size=(150, 150)) # this is a PIL image x = img_to_array(img) # Numpy array with shape (150, 150, 3) x = x.reshape((1,) + x.shape) # Numpy array with shape (1, 150, 150, 3) # Rescale by 1/255 x /= 255 # Let's run our image through our network, thus obtaining all # intermediate representations for this image. successive_feature_maps = visualization_model.predict(x) # These are the names of the layers, so can have them as part of our plot layer_names = [layer.name for layer in model.layers] # Now let's display our representations for layer_name, feature_map in zip(layer_names, successive_feature_maps):
#finetune you network parameter in last by using low learning rate like 0.00001
model.save('hyperparameter1.h5')
#from tensorflow.keras.models import load_model
#model = load_model('LETSDOTHIS.h5')
from tensorflow.keras.preprocessing import image
import matplotlib.pyplot as plt
import numpy as np
results = []
test = pd.read_csv('test.csv')
length_test = len(test['Image_File'])
testDF = pd.read_csv('test.csv')
for i in range(length_test):
img1 = image.load_img('DataSet/Test Images/' + test['Image_File'][i],
target_size=(720, 480))
img = image.img_to_array(img1)
img = img / 255
# create a batch of size 1 [N,H,W,C]
img = np.expand_dims(img, axis=0)
prediction = model.predict(img, batch_size=None,
steps=1) # gives all class prob.
if (prediction[:, :] > 0.5):
testDF['Class'][i] = "Small"
else:
testDF['Class'][i] = "Large"
print(i)
print(testDF)
testDF.to_csv('hyperparameterTWEAK.csv')
# construct the argument parse and parse the arguments
ap = argparse.ArgumentParser()
ap.add_argument("-i", "--image", required=True,
help="path to the input image")
ap.add_argument("-o", "--output", required=True,
help="path to output directory to store augmentation examples")
ap.add_argument("-p", "--prefix", type=str, default="image",
help="output filename prefix")
args = vars(ap.parse_args())
# load the input image, convert it to a NumPy array, and then
# reshape it to have an extra dimension
print("[INFO] loading example image...")
image = load_img(args["image"])
image = img_to_array(image)
image = np.expand_dims(image, axis=0)
# construct the image generator for data augmentation then
# initialize the total number of images generated thus far
aug = ImageDataGenerator(rotation_range=30, width_shift_range=0.1,
height_shift_range=0.1, shear_range=0.2, zoom_range=0.2,
horizontal_flip=True, fill_mode="nearest")
total = 0
# construct the actual Python generator
print("[INFO] generating images...")
imageGen = aug.flow(image, batch_size=1, save_to_dir=args["output"],
save_prefix=args["prefix"], save_format="jpg")
# loop over examples from our image data augmentation generator
def ModelKi(location,BASE_DIR):
print("first:" + location)
base_model = InceptionV3(weights='imagenet', include_top=False,input_shape=(150,150,3))
#base_model = InceptionV3(include_top=False, weights='imagenet',input_shape=(150,150,3))
temploc = location.split('\\')
print("------------------------|||||||||||")
print(temploc)
baseloc = ''
for i in range(len(temploc)-1):
baseloc += temploc[i] + '\\'
print("third:" + baseloc)
#print("fourth:"+location)
#defining the model architecture
model = Sequential()
model.add(Dense(1024, activation='relu',input_shape=(18432,)))
model.add(Dropout(0.5))
model.add(Dense(512, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(256, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(128, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(101, activation='softmax'))
model.load_weights(BASE_DIR + "\DV\models\weight_150_300_Incv3.hdf5")
model.compile(loss='categorical_crossentropy',optimizer='Adam',metrics=['accuracy'])
# creating the tags
train = pd.read_csv(BASE_DIR+'\DV\\train_new.csv')
y = train['class']
y = pd.get_dummies(y)
pred = []
cap = cv2.VideoCapture(location)
print(cap)# capturing the video from the given path
frameRate = cap.get(5) #frame rate
x=1
count = 0
while(cap.isOpened()):
frameId = cap.get(1) #current frame number
ret, frame = cap.read()
if (ret != True):
break
if (frameId % math.floor(frameRate) == 0):
# storing the frames of this particular video in temp folder
filename = str(baseloc) + 'temp\\' + "_frame%d.jpg" % count;count+=1
cv2.imwrite(filename, frame)
cap.release()
images = glob(str(baseloc) + "temp\*.jpg")
prediction_images = []
for i in range(len(images)):
img = image.load_img(images[i], target_size=(150,150,3))
img = image.img_to_array(img)
img = img/255
prediction_images.append(img)
# converting all the frames for a test video into numpy array
prediction_images = np.array(prediction_images)
# extracting features using pre-trained model
prediction_images = base_model.predict(prediction_images)
# converting features in one dimensional array
prediction_images = prediction_images.reshape(prediction_images.shape[0], 3*3*2048)
# predicting tags for each array
predictionar = model.predict_classes(prediction_images)
predictionar = predictionar.reshape(predictionar.shape[0],1)
mode = s.mode(predictionar)
x = mode[0][0]
ans = y.columns.values[x][0]
files = glob(str(baseloc) + 'temp\*')
for f in files:
os.remove(f)
return ans
csv_file = os.path.join(test_path, 'test.csv')
df = pd.read_csv(csv_file)
print(df.head())
submission = './submission.csv'
csv_sub = open(submission, mode='w')
writer = csv.writer(csv_sub,
delimiter=',',
quotechar='"',
quoting=csv.QUOTE_MINIMAL)
writer.writerow(['Image', 'target'])
dances = [
'bharatanatyam', 'kathak', 'kathakali', 'kuchipudi', 'manipuri',
'mohiniyattam', 'odissi', 'sattriya'
]
for index, row in df.iterrows():
print("**************>", row.values[0])
test_image = os.path.join(test_image_path, row.values[0])
img = image.load_img(test_image, target_size=(image_width, image_height))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
images = np.vstack([x])
classes = model.predict(images, batch_size=10)
image_class = np.argmax(classes[0])
result = [row.values[0], dances[image_class]]
writer.writerow(result)
csv_sub.close()
test_covid_path = glob('.\\data-split\\val\\CT_COVID\\*')
test_normal_path = glob('.\\data-split\\val\\CT_NonCOVID\\*')
# Load images (For google collab)
# train_covid_path = glob('/content/data-split/train/CT_COVID/*')
# train_normal_path = glob('/content/data-split/train/CT_NonCOVID/*')
# test_covid_path = glob('/content/data-split/val/CT_COVID/*')
# test_normal_path = glob('/content/data-split/val/CT_NonCOVID/*')
# Building the training array
cnt = 0 # index
for img_file in train_covid_path:
image_orig = cv2.imread(img_file)
image_resized = cv2.resize(image_orig, (256, 256),
interpolation=cv2.INTER_CUBIC)
img_array = img_to_array(image_resized)
x_train_temp[cnt] = img_array / 255
y_train_temp[cnt] = 1
cnt += 1
for img_file in train_normal_path:
image_orig = cv2.imread(img_file)
image_resized = cv2.resize(image_orig, (256, 256),
interpolation=cv2.INTER_CUBIC)
img_array = img_to_array(image_resized)
x_train_temp[cnt] = img_array / 255
y_train_temp[cnt] = 0
cnt += 1
def path_to_tensor(img_path):
img = image.load_img(img_path, target_size=(299, 299))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
return x
# In[3]:
pathname = r"C:\Users\LENOVO\Desktop\Face-Mask-Detection\dataset"
print("[INFO] loading images...")
imagePaths = list(paths.list_images(pathname))
data = []
labels = []
# In[4]:
for imagePath in imagePaths:
# extract the class label from the filename
label = imagePath.split(os.path.sep)[-2]
# load the input image (224x224) and preprocess it
image = load_img(imagePath, target_size=(224, 224))
image = img_to_array(image)
image = preprocess_input(image)
# update the data and labels lists, respectively
data.append(image)
labels.append(label)
# convert the data and labels to NumPy arrays
data = np.array(data, dtype="float32")
labels = np.array(labels)
# perform one-hot encoding on the labels
lb = LabelBinarizer()
labels = lb.fit_transform(labels)
labels = to_categorical(labels)
