class PriusPredictor(object):
def __init__(self, image_path, model_path, output_path):
self.avgColor = []
self.pcaColors = []
self.detector = ObjectDetection()
self.detector.setModelTypeAsYOLOv3()
self.detector.setModelPath(model_path + "yolo.h5")
self.detector.loadModel(detection_speed="flash")
self.prediction = CustomImagePrediction()
self.prediction.setModelTypeAsResNet()
#self.prediction.setModelPath(model_path + "model_ex-012_acc-0.988819.h5")
self.prediction.setModelPath(model_path + "model_ex-043_acc-0.996787.h5")
self.prediction.setJsonPath(model_path + "model_class.json")
self.prediction.loadModel(num_objects=2)
now = time.localtime()
self.frame_folder = str(now.tm_year) + str(now.tm_mon) + str(now.tm_mday)
self.image_path = image_path
self.output_path = output_path + "detection/" + self.frame_folder + "/"
if os.path.exists(image_path) is False:
os.mkdir(image_path)
if os.path.exists(output_path) is False:
os.mkdir(output_path)
if os.path.exists(os.path.join(output_path, 'detection')) is False:
os.mkdir(os.path.join(output_path, 'detection'))
if os.path.exists(os.path.join(output_path, 'processed')) is False:
os.mkdir(os.path.join(output_path, 'processed'))
self.create_output_folder()
def create_output_folder(self):
if os.path.exists(self.output_path) is False:
os.mkdir(self.output_path)
def predict_vehicle_method(self, prediction_meta):
detected_img = os.path.join(prediction_meta['image_path'], prediction_meta['image_name'])
if os.path.exists(detected_img) is not True:
detected_img = prediction_meta['image_path']
return self.prediction.predictImage(detected_img, result_count=2)
def predict_vehicle(self, prediction_meta):
detected_img = prediction_meta['image_path']
return self.prediction.predictImage(detected_img, result_count=2)
def detect_pca(self, image):
priusImage = PriusImage.from_path(image)
return priusImage.has_pca_match()
def detect_vehicle(self, meta_data):
try:
image = os.path.join(meta_data["image_path"], meta_data['image_name'])
output_image = self.output_path + meta_data['image_name']
#print("Detecting vehicle for " + meta_data['image_name'] + " -> " + output_image)
if os.path.exists(image) is not True:
print("File doesnt exist. File: " + image)
custom_objects = self.detector.CustomObjects(car=True)
detections, objects_path = self.detector.detectCustomObjectsFromImage(custom_objects=custom_objects,
input_image=image,
extract_detected_objects=True,
output_image_path=output_image,
minimum_percentage_probability=50)
return zip(detections, objects_path)
except Exception as e:
print("While detecting vehicle: " + str(e))
def detect_vehicle_from_array(self, decoded):
#print("Detecting vehicle for " + meta_data['image_name'] + " -> " + output_image)
custom_objects = self.detector.CustomObjects(car=True)
result = self.detector.detectCustomObjectsFromImage(custom_objects=custom_objects,
input_type="array",
input_image=np.array(decoded),
#output_type="array",
minimum_percentage_probability=50)
print("Detected: " + str(result))
return result
textSizeWidth, textSizeHeight = textSize
ptLowerLeftTextOriginX = int(ptCenterOfTextAreaX - (textSizeWidth / 2))
ptLowerLeftTextOriginY = int(ptCenterOfTextAreaY + (textSizeHeight / 2))
cv2.putText(imgOriginalScene, licPlate.strChars,
(ptLowerLeftTextOriginX, ptLowerLeftTextOriginY), intFontFace,
fltFontScale, SCALAR_YELLOW, intFontThickness)
execution_path = os.getcwd()
detector = ObjectDetection()
detector.setModelTypeAsRetinaNet()
detector.setModelPath(
os.path.join(execution_path, "resnet50_coco_best_v2.0.1.h5"))
detector.loadModel()
# cv2.imwrite("imgg.jpg", frames)
# cv2.imwrite("imgg.jpg",frames)
detections = detector.detectObjectsFromImage(
input_image=os.path.join(execution_path, "shah.jpg"),
output_image_path=os.path.join(execution_path, "imagenew.jpg"))
for eachObject in detections:
if eachObject["name"] == "car":
# cv2.imshow("Image",eachObject["box_image"])
number_plate(eachObject["box_image"])
# print(eachObject["color"])
# save_image = Image.fromarray(eachObject["box_image"],'RGB')
# save_image.save('my.png')
# break
# cv2.imwrite()
# save_image = misc.imread(eachObject["box_image"])
class LiveDetector:
def __init__(self):
# Instantiate detector
self.detector = ObjectDetection()
# Set and load model
self.model_path = "D:/Final-Year-Project/Object-tracking/models/yolo.h5"
self.detector.setModelTypeAsYOLOv3()
self.detector.setModelPath(self.model_path)
self.detector.loadModel()
# Set custom objects
self.custom_objects = self.detector.CustomObjects(car=True,
motorcycle=True,
person=True,
bicycle=True,
dog=True)
self.tracker = CentroidTracker()
def track_objects(self, frame):
rects = []
names = []
data = {}
frame = self.pixelate_frontyard((100, 100), frame)
returned_image, detection = self.detector.detectCustomObjectsFromImage(
custom_objects=self.custom_objects,
input_image=frame,
output_type="array",
input_type="array")
for eachObject in detection:
rects.append(eachObject["box_points"])
names.append(eachObject["name"])
(startX, startY, endX, endY) = eachObject["box_points"]
cv2.rectangle(frame, (startX, startY), (endX, endY), (0, 255, 0),
2)
self.blur_object((startX, startY), (endX, endY), (11, 11), frame)
objects = self.tracker.update(rects, names)
if objects is not None:
for objectID, objectDetails in objects.items():
# draw both the ID of the object and the centroid of the
# object on the output frame
centroid = objectDetails[0]
name = objectDetails[1]
if self.tracker.disappeared[objectID] < 1:
text = name + " " + str(objectID)
cv2.putText(frame, text,
(centroid[0] - 30, centroid[1] - 10),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)
cv2.circle(frame, (centroid[0], centroid[1]), 4,
(0, 255, 0), -1)
data[name] = objectID
return frame, data
def blur_object(self, topLeft, bottomRight, kSize, frame):
x, y = topLeft[0], topLeft[1]
w, h = bottomRight[0] - topLeft[0], bottomRight[1] - topLeft[1]
ROI = frame[y:y + h, x:x + w]
blur = cv2.GaussianBlur(ROI, kSize, 0)
frame[y:y + h, x:x + w] = blur
def blur_frontyard(self, kSize, frame):
height, width, channel = frame.shape
ROI_corners = np.array([[(320, 490), (895, 320), (895, height),
(320, height)]],
dtype=np.int32)
blurred_frame = cv2.GaussianBlur(frame, kSize, 0)
mask = np.zeros(frame.shape, dtype=np.uint8)
ignore_mask_color = (255, ) * channel
cv2.fillPoly(mask, ROI_corners, ignore_mask_color)
mask_inverse = np.ones(mask.shape).astype(np.uint8) * 255 - mask
frame = cv2.bitwise_and(blurred_frame, mask) + cv2.bitwise_and(
frame, mask_inverse)
return frame
def pixelate_frontyard(self, kSize, frame):
height, width, channel = frame.shape
w, h = kSize
ROI_corners = np.array([[(320, 490), (895, 320), (895, height),
(320, height)]],
dtype=np.int32)
temp = cv2.resize(frame, (w, h), interpolation=cv2.INTER_LINEAR)
pixelated_frame = cv2.resize(temp, (width, height),
interpolation=cv2.INTER_NEAREST)
mask = np.zeros(frame.shape, dtype=np.uint8)
ignore_mask_color = (255, ) * channel
cv2.fillPoly(mask, ROI_corners, ignore_mask_color)
mask_inverse = np.ones(mask.shape).astype(np.uint8) * 255 - mask
frame = cv2.bitwise_and(pixelated_frame, mask) + cv2.bitwise_and(
frame, mask_inverse)
return frame
import os
import cv2
from PIL import Image
cap = cv2.VideoCapture(0)
# Load Yolo
execution_path = os.getcwd()
detector = ObjectDetection()
#detector.setModelTypeAsTinyYOLOv3() # YOLOv3
#detector.setModelPath( os.path.join(execution_path , "models/yolo-tiny.h5"))
detector.setModelTypeAsRetinaNet() # Other types are TinyYOLOv3, YOLOv3
detector.setModelPath( os.path.join(execution_path , "models/resnet50_coco_best_v2.0.1.h5")) #yolo-tiny.h5
detector.loadModel() #detection_speed="fastest"
process_this_frame = True
while 1:
# reads frames from a camera
ret, img = cap.read()
# img_str = cv2.imencode('.jpg', img)[1].tostring()
if process_this_frame:
img = cv2.resize(img, (0,0), fx=0.5, fy=0.5)
detections = detector.detectObjectsFromImage(input_type="array", input_image=img, output_type="file")
print(detections)
def save_result(result):
with open("prius_results.json") as json_file:
data = json.load(json_file)
temp = data
temp.append(result)
write_json(data)
detector = ObjectDetection()
detector.setModelTypeAsYOLOv3()
detector.setModelPath("./yolo.h5")
detector.loadModel(detection_speed="fastest")
prediction = CustomImagePrediction()
prediction.setModelTypeAsResNet()
# self.prediction.setModelPath(model_path + "model_ex-012_acc-0.988819.h5")
prediction.setModelPath("./model_ex-043_acc-0.996787.h5")
prediction.setJsonPath("./model_class.json")
prediction.loadModel(prediction_speed="fastest")
custom_objects = detector.CustomObjects(car=True)
accuracy = 0
imgs = []
data = requests.get("http://seattle.gov/trafficcams/images/15_NW_65_NS.jpg").content
decoded = cv2.imdecode(np.frombuffer(data, np.uint8), -1)
# class_text = ['run','sit','stand','walk','standup']
class_text = [
'a01', 'a02', 'Two Hand Wave', 'a04', 'a05', 'a06', 'a07', 'a08', 'a09',
'a10', 'a11', 'a12', 'hand clap', 'a14', 'a15', 'a16', 'a17', 'a18'
]
### ImageAI
# model_path = "pretrain/yolo-tiny.h5"
model_path = "pretrain/yolo.h5"
detector = ObjectDetection()
# detector.setModelTypeAsTinyYOLOv3()
detector.setModelTypeAsYOLOv3()
detector.setModelPath(model_path)
detector.loadModel(
detection_speed="normal"
) #"normal"(default), "fast", "faster" , "fastest" and "flash".
custom_objects = detector.CustomObjects(person=True)
### Main
cap = cv2.VideoCapture(0)
# Crop setting
width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)) # float
height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)) # float
length_x_q = [[] for i in range(max_person)]
length_y_q = [[] for i in range(max_person)]
max_length = 5
mid_x_q = [[] for i in range(max_person)]
mid_y_q = [[] for i in range(max_person)]
max_mid = 3
def object(self, selection):
if selection == []:
layout = BoxLayout(orientation="vertical", spacing=10, padding=10)
popupLabel = Label(text="No file is selected .")
print("No file is selected .")
closeButton = Button(background_normal='img\close.png', background_down='img\close.png',
size_hint=(.38, .6), pos_hint={"center_x": .5})
layout.add_widget(popupLabel)
layout.add_widget(closeButton)
popup = Popup(title='Error', content=layout, size_hint=(.8, .6))
popup.open()
closeButton.bind(on_press=popup.dismiss)
engine = pyttsx3.init()
print("No file is selected .")
if self.soundcount == 0 :
engine.say("No file is selected .")
engine.runAndWait()
else :
engine = pyttsx3.init()
print("Please Wait .")
if self.soundcount == 0:
engine.say("Please Wait .")
engine.runAndWait()
detector = ObjectDetection()
model_path = "./model/yolo-tiny.h5"
input_path = selection[0]
output_path = "{} obj.png".format(selection[0])
detector.setModelTypeAsTinyYOLOv3()
detector.setModelPath(model_path)
detector.loadModel()
detection = detector.detectObjectsFromImage(input_image=input_path, output_image_path=output_path)
engine = pyttsx3.init()
print("Input Path : {}".format(input_path))
print("Output Path : {}".format(output_path))
if self.soundcount == 0:
engine.say("There are ")
print("There are ")
flag = 0
temp=""
items=""
for eachItem in detection:
print(eachItem["name"], " : ", eachItem["percentage_probability"])
if self.soundcount == 0:
if eachItem["name"] != temp:
engine.say(eachItem["name"])
temp = eachItem["name"]
items = items + temp + ", "
flag = flag + 1
if flag == 0:
if self.soundcount == 0:
engine.say("no object's in the image .")
print("no object's in the image .")
else:
if self.soundcount == 0:
engine.say("in the image .")
print("in the image .")
layout = BoxLayout(orientation="vertical", spacing=10, padding=10)
popupImage = Image(source = output_path)
closeButton = Button(background_normal='img\close.png', background_down='img\close.png', size_hint=(.38, .31),
pos_hint={"center_x": .5})
layout.add_widget(popupImage)
layout.add_widget(closeButton)
popup = Popup(title=items, content=layout, size_hint=(.8, .9))
popup.open()
closeButton.bind(on_press=popup.dismiss)
engine.runAndWait()
def object_detection():
execution_path = os.getcwd()
detector = ObjectDetection()
detector.setModelTypeAsRetinaNet()
detector.setModelPath(settings.OBJECT_RECOGNITION_MODEL)
detector.loadModel()
done = 0
total_photos = Photo.objects.count()
t = datetime.now()
for photo in Photo.objects.all().order_by('pk'):
if not photo.exists(check_input=True):
logger.warning("Path does not exist!", id=photo.id, name=photo.name)
total_photos -= 1
continue
if photo.get_objects():
total_photos -= 1
continue
ext = photo.name.split('.')[1]
tmp_file = os.path.join(execution_path, f"photo.{ext}")
new_tmp_file = os.path.join(execution_path, f"photo_out.{ext}")
if os.path.islink(tmp_file):
os.unlink(tmp_file)
os.symlink(photo.get_input_path(), tmp_file,)
detections = detector.detectObjectsFromImage(
input_image=tmp_file,
output_image_path=new_tmp_file
)
with transaction.atomic():
for detection in detections:
object_category, _ = ObjectCategory.objects.get_or_create(name=detection["name"])
ObjectIdentification.objects.create(
photo=photo,
category=object_category,
confidence=detection["percentage_probability"]
)
os.unlink(tmp_file)
if os.path.isfile(photo.get_output_path()):
os.remove(photo.get_output_path())
os.rename(new_tmp_file, photo.get_output_path())
done += 1
elapsed_time = (datetime.now() - t).seconds
estimated_time_left = elapsed_time / done * (total_photos - done)
stats = dict(
name=photo.name,
id=photo.pk,
done=done,
left=total_photos-done,
proc=f'{"%.2f" % (done / total_photos * 100)}',
est_left=f'{"%.2f" % estimated_time_left}s'
)
if not len(detections):
logger.warning(
"No objects found!",
**stats
)
else:
logger.info(
"Detecting",
**stats
)
logger.info("Finished image recognition")
def main(origin_images_path="../data/raw_data/cars_train",
destination_images_path="../data/object_detection_data/output_images_YOLO",
final_images_path="../data/object_detection_data", sample=True):
file_path = (os.path.dirname(os.path.abspath(__file__))).replace('\\', '/')
if not os.path.exists(destination_images_path):
os.makedirs(destination_images_path)
logging.info('Starting detecting objects')
detector = ObjectDetection()
detector.setModelTypeAsYOLOv3()
detector.setModelPath("../data/raw_data/YOLO_weights/yolo.h5")
detector.loadModel()
images = os.listdir(origin_images_path)
if sample:
images = images[:10]
for i in images:
detector.detectObjectsFromImage(input_image=origin_images_path + '/' + i,
output_image_path=destination_images_path + "/new_{}".format(i),
extract_detected_objects=True)
logging.info('Finished detecting objects')
logging.info('Starting assigining objects to folder output_image_cropped')
# Keep only the biggest cut car
dirs = list(filter(os.path.isdir, [destination_images_path + '/' + i for i in os.listdir(destination_images_path)]))
for directory in dirs:
files = os.listdir(directory)
cars_size = {}
for file in files:
if not (file.startswith('car') or file.startswith('truck')):
os.unlink(directory + "/" + file)
remaining_files = os.listdir(directory)
for file in remaining_files:
cars_size[file] = Image.open(directory + str("/") + file).size
if len(cars_size) > 1:
biggest_car = None
dim = (0, 0)
for car in cars_size.keys():
if cars_size[car][0] * cars_size[car][1] > dim[0] * dim[1]:
biggest_car = car
dim = cars_size[car]
to_delete = (list(set(cars_size.keys())))
to_delete.remove(biggest_car)
for small_car in to_delete:
os.unlink(directory + str("/") + small_car)
# Rename the images as number.jpg
dirs = filter(os.path.isdir, [destination_images_path + '/' + i for i in os.listdir(destination_images_path)])
for directory in dirs:
files = os.listdir(directory)
for file in files:
number = re.search(r"[0-9]+", str(directory))
new_name = str(number.group())+".jpg"
start = directory + str("/") + file
end = directory + str("/") + new_name
os.rename(start, end)
# Put the all the cut cars into a folder named "output_images_cropped"
dirs = filter(os.path.isdir, [destination_images_path + '/' + i for i in os.listdir(destination_images_path)])
if not os.path.exists(final_images_path + "/output_images_cropped"):
os.mkdir(final_images_path + "/output_images_cropped")
for directory in dirs:
files = os.listdir(directory)
for file in files:
start = directory + str("/") + file
destination = str(final_images_path + "/output_images_cropped") + str("/") + file
shutil.copyfile(start, destination)
logging.info('Finished entire process')
class ObjectOnlyCompressor:
def __init__(self, model_path, api_key):
self.detector = ObjectDetection()
self.detector.setModelTypeAsRetinaNet()
self.detector.setModelPath(model_path)
self.detector.loadModel()
self.api_key = api_key
def compress(self, input_path, output_path):
image = Image.open(input_path).convert('RGB')
image.thumbnail((255, 255))
items = self.detector.detectObjectsFromImage(
input_image=np.asarray(image),
input_type='array',
output_image_path='debug.jpg',
)
file_content = bytearray()
for item in items:
file_content.extend(item['name'].encode('ascii'))
file_content.append(0)
file_content.extend(item['box_points'])
with open(output_path, 'wb') as file:
file.write(file_content)
def decompress(self, input_path, output_path):
with open(input_path, 'rb') as input_file:
input_file_contents = input_file.read()
output_image = Image.new('RGB', (255, 255), (255, 255, 255))
items = []
while input_file_contents:
delimiter_index = input_file_contents.index(0)
name = input_file_contents[:delimiter_index].decode('ascii')
box_points = [
int(input_file_contents[delimiter_index + offset])
for offset in range(1, 5)
]
items.append((name, box_points))
input_file_contents = input_file_contents[delimiter_index + 5:]
for item in items:
name, box_points = item
print(name, box_points)
response = requests.post("https://api.deepai.org/api/text2img",
data={'text': name},
headers={'api-key': self.api_key})
url = response.json()['output_url']
image = fetch_image(url)
image = image.resize((
box_points[2] - box_points[0],
box_points[3] - box_points[1],
))
output_image.paste(image, (box_points[0], box_points[1]))
output_image.save(output_path)
class App:
def __init__(self, master, portV):
# initiating Arduino serial port
if portV:
try:
self.arduino = serial.Serial(portV, 9600)
except (serial.SerialException):
tk.messagebox.showinfo("Error", "Invalid port.")
else:
self.arduino = None
self.cam = cv2.VideoCapture(1)
self.lightMode = 0
self.followMode = True
self.height = 0
self.width = 0
self.xBound = 0
self.yBound = 0
self.x = 0
self.y = 0
# Initiating object detection model
self.detector = ObjectDetection()
self.detector.setModelTypeAsYOLOv3()
self.detector.setModelPath("yolo.h5")
self.custom = self.detector.CustomObjects(person=True)
self.detector.loadModel("faster")
self.start = 0
# Creating GUI
self.master = master
master.title('Feed')
self.cFont = font.Font(family='Comic Sans MS',
size=15,
weight=font.BOLD)
frame = tk.Frame(master, bg='white')
frame.pack()
self.timeLabel = tk.Label(frame,
text="FPS: ",
fg='white',
bg='orange',
width=15,
height=1)
self.timeLabel['font'] = self.cFont
if showFeed:
self.timeLabel.grid(row=0, column=0)
else:
self.timeLabel.grid(row=0, column=1)
self.lightToggle = tk.Button(frame,
text="Light:\nOff",
fg='white',
bg='orange',
height=3,
width=15,
cursor='arrow',
command=self.toggleLight)
self.lightToggle['font'] = self.cFont
if showFeed:
self.lightToggle.grid(row=2, column=0, sticky='W')
else:
self.lightToggle.grid(row=1, column=0)
self.followToggle = tk.Button(frame,
text="Following:\nTrue",
fg='white',
bg='orange',
height=3,
width=15,
cursor='arrow',
command=self.toggleFollow)
self.followToggle['font'] = self.cFont
if showFeed:
self.followToggle.grid(row=2, column=0)
else:
self.followToggle.grid(row=1, column=1)
self.stop = tk.Button(master=frame,
text="Stop",
fg='white',
bg='orange',
height=3,
width=15,
cursor='arrow',
command=self.master.destroy)
self.stop['font'] = self.cFont
if showFeed:
self.stop.grid(row=2, column=0, sticky='E')
else:
self.stop.grid(row=1, column=2)
ret, output = self.cam.read()
# Set image bounds
self.height, self.width, channels = output.shape
self.xBound = self.width / 10
self.yBound = self.height / 8
detPic, det = self.detector.detectCustomObjectsFromImage(
custom_objects=self.custom,
input_type="array",
input_image=output,
output_type="array",
minimum_percentage_probability=15)
if showFeed:
imageN = ImageTk.PhotoImage(image=Image.fromarray(detPic))
self.imgLabel = tk.Label(frame, image=imageN)
self.imgLabel.grid(row=1, column=0)
self.loop()
# Turn light on and off
def toggleLight(self):
if self.lightMode == 0:
self.lightMode = 1
self.lightToggle.configure(text="Light:\nOn")
else:
self.lightMode = 0
self.lightToggle.configure(text="Light:\nOff")
# Turn object detection and following on/off
def toggleFollow(self):
self.followMode = not self.followMode
if self.followMode:
self.followToggle.configure(text="Following:\nTrue")
else:
self.followToggle.configure(text="Following:\nFalse")
def loop(self):
self.start = time.time()
ret, output = self.cam.read()
# If follow mode is set to on
if self.followMode:
# Detect hands
detPic, det = self.detector.detectCustomObjectsFromImage(
custom_objects=self.custom,
input_type="array",
input_image=output,
output_type="array",
minimum_percentage_probability=15)
# If hand is detected
if len(det) > 0:
#Get location of detected objects
location = det[0]['box_points']
self.x = (location[0] + location[2]) / 2
self.y = (location[1] + location[3]) / 2
if (self.x > (self.width / 2 + self.xBound)):
self.x = 2
elif (self.x < (self.width / 2 - self.xBound)):
self.x = 0
else:
self.x = 1
if (self.y > (self.height / 2 + self.yBound)):
self.y = 0
elif (self.y < (self.height / 2 - self.yBound)):
self.y = 2
else:
self.y = 1
else:
self.x = 1
self.y = 1
print(str(self.x) + " " + str(self.y) + " " + str(self.lightMode))
# Send to Arduino over serial
if self.arduino is not None:
self.arduino.write(
struct.pack('>BBB', self.x, self.y, self.lightMode))
else:
# If nothing is detected
concatted = "1 1 " + str(self.lightMode)
print(concatted)
if self.arduino is not None:
self.arduino.write(struct.pack('>BBB', 1, 1, self.lightMode))
if showFeed:
# Display image
if self.followMode:
imageN = ImageTk.PhotoImage(image=Image.fromarray(detPic))
else:
imageN = ImageTk.PhotoImage(image=Image.fromarray(output))
self.imgLabel.configure(image=imageN)
self.imgLabel.image = imageN
# FPS counter
self.timeLabel.configure(text="FPS: " +
str(1 / (time.time() - self.start))[0:5])
else:
if self.followMode:
self.timeLabel.configure(
text="FPS: " + str(1 / (time.time() - self.start))[0:5])
else:
self.timeLabel.configure(text="FPS: Very Fast")
# Loop back to itself
self.master.after(1, self.loop)
def predict():
if request.method == 'POST':
from keras import backend as K
K.clear_session()
print("POST")
# loading model here in background
detector = ObjectDetection()
# use fat yolo since tiny yolo sux
detector.setModelTypeAsYOLOv3()
path = "./models/yolo.h5"
detector.setModelPath(path)
detector.loadModel()
# creating list of custom objects
custom = detector.CustomObjects(backpack=True, umbrella=True, cup=True)
object_plastic_map = plastic_dict.plastic_dict
# req_data = request.get_json()
# image = req_data["image"]
print(list(request.form.keys()))
str_image = request.form["image"]
print(str_image[:20])
str_image = base64.b64decode(str_image)
print(str_image[:20])
# str_image = str_image.decode('utf-8')
# str_image = base64_bytes.decode("base64")
# print(str_image)
# converting bytes information into string into np array
# str_image = image.decode("utf-8")
# print(str_image)
image = Image.open(BytesIO(str_image)).convert("RGB")
# image = ast.literal_eval(str_image)
# print(image)
im_array = np.array(image,dtype=np.uint8)
# print(np.shape(image))
print(np.shape(im_array))
detection = detector.detectCustomObjectsFromImage(custom_objects=custom,
input_type="array", input_image=im_array,
output_type="array",
minimum_percentage_probability=70)
K.clear_session()
to_return = ""
for eachItem in detection[1]:
name = eachItem["name"]
print(name + " : ", eachItem["percentage_probability"])
to_return += name
# to_return += ":"
# to_return += str(object_plastic_map[name])
to_return += str(",")
if to_return == "":
print("nothing detected")
print(to_return)
return to_return
else:
return no_input()
# In[2]:
model_path = os.getcwd()
# In[3]:
PRE_TRAINED_MODELS = ["yolo.h5"]
# In[4]:
# create imageAI objects and load models
object_detector = ObjectDetection()
object_detector.setModelTypeAsYOLOv3()
object_detector.setModelPath(os.path.join(model_path, PRE_TRAINED_MODELS[0]))
object_detector.loadModel()
object_detections = object_detector.detectObjectsFromImage(
input_image="people_umbrella.jpg")
# In[5]:
#define model paths and allow file extention
UPLOAD_FOLDER = model_path
ALLOWED_EXTENSIONS = set(['png', 'jpg', 'jpeg', 'gif'])
# In[6]:
app = Flask(__name__)
app.config['UPLOAD_FOLDER'] = UPLOAD_FOLDER
# In[7]:
def main():
# setup folders
image_path = "../kaggle/challenge2018_test/"
model_weight_path = "resnet50_coco_best_v2.0.1.h5"
print('Reading and transformin class information ...')
# Read classes from RetinaNet. Prepared list taken from: https://github.com/fizyr/keras-retinanet/blob/master/examples/ResNet50RetinaNet.ipynb
labels_to_names = {
0: 'person',
1: 'bicycle',
2: 'car',
3: 'motorcycle',
4: 'airplane',
5: 'bus',
6: 'train',
7: 'truck',
8: 'boat',
9: 'traffic light',
10: 'fire hydrant',
11: 'stop sign',
12: 'parking meter',
13: 'bench',
14: 'bird',
15: 'cat',
16: 'dog',
17: 'horse',
18: 'sheep',
19: 'cow',
20: 'elephant',
21: 'bear',
22: 'zebra',
23: 'giraffe',
24: 'backpack',
25: 'umbrella',
26: 'handbag',
27: 'tie',
28: 'suitcase',
29: 'frisbee',
30: 'skis',
31: 'snowboard',
32: 'sports ball',
33: 'kite',
34: 'baseball bat',
35: 'baseball glove',
36: 'skateboard',
37: 'surfboard',
38: 'tennis racket',
39: 'bottle',
40: 'wine glass',
41: 'cup',
42: 'fork',
43: 'knife',
44: 'spoon',
45: 'bowl',
46: 'banana',
47: 'apple',
48: 'sandwich',
49: 'orange',
50: 'broccoli',
51: 'carrot',
52: 'hot dog',
53: 'pizza',
54: 'donut',
55: 'cake',
56: 'chair',
57: 'couch',
58: 'potted plant',
59: 'bed',
60: 'dining table',
61: 'toilet',
62: 'tv',
63: 'laptop',
64: 'mouse',
65: 'remote',
66: 'keyboard',
67: 'cell phone',
68: 'microwave',
69: 'oven',
70: 'toaster',
71: 'sink',
72: 'refrigerator',
73: 'book',
74: 'clock',
75: 'vase',
76: 'scissors',
77: 'teddy bear',
78: 'hair drier',
79: 'toothbrush'
}
retina_classes = pd.DataFrame(list(labels_to_names.items()),
columns=['ID', 'Label'])
# Read classes from Google
google_classes = pd.read_csv("class-descriptions-boxable.csv",
delimiter=";",
names=['ID', 'Label'])
google_classes['Label'] = google_classes['Label'].apply(str.lower)
# Iterate over Retina_classes and check if they're in Google classes.
# If so, take google ID in new class mappings
label_mappings = pd.DataFrame(columns=['ID', 'Label'])
coco_to_google = {}
for i in range(0, retina_classes.shape[0]):
for j in range(0, google_classes.shape[0]):
if (retina_classes["Label"][i] == google_classes["Label"][j]):
label_mappings.loc[i] = google_classes.iloc[j]
coco_to_google[retina_classes["Label"]
[i]] = google_classes.iloc[j]['ID']
print('Loading resnet model ...')
model_weight_path = "resnet50_coco_best_v2.0.1.h5"
detector = ObjectDetection()
detector.setModelTypeAsRetinaNet()
detector.setModelPath(model_weight_path)
detector.loadModel()
print('Starting object detection method ...')
detect_objects(detector_fn=detector.detectObjectsFromImage,
img_path=image_path,
results_fname='challenge_submission.csv',
minimum_percentage_probability=50,
translation_dict=coco_to_google)
class CaptionDataset():
def __init__(self):
self.execution_path = os.getcwd()
self.detector = ObjectDetection()
self.detector.setModelTypeAsYOLOv3()
self.detector.setModelPath(os.path.join(self.execution_path,
"yolo.h5"))
self.detector.loadModel(detection_speed="flash")
self.transforms = tv.transforms.Compose(
[tv.transforms.Resize(224),
tv.transforms.ToTensor(), normalize])
data = t.load('caption.pth')
self.ix2id = data['ix2id']
img_path = 'ai_challenger_caption_train_20170902/caption_train_images_20170902/'
imgs = [os.path.join(img_path, self.ix2id[_]) \
for _ in range(len(self.ix2id))]
self.imgs = imgs
def __getitem__(self, index):
img = Image.open(self.imgs[index]).convert('RGB')
img_area = img.size[0] * img.size[1]
detections = self.detector.detectObjectsFromImage(
input_image=self.imgs[index])
block_num = len(detections)
cut_imgs = []
if block_num == 2:
for eachObject in detections:
cropped = img.crop(eachObject["box_points"])
# print(eachObject["box_points"])
m = eachObject["box_points"][2] - eachObject["box_points"][0]
n = eachObject["box_points"][3] - eachObject["box_points"][1]
# print('截图宽：%d,截图高：%d' % (m, n))
jie = (m * n) / img_area * 100
# print('截图占比：%.2f%%' % jie)
if jie < 10:
cut_imgs.append(self.transforms(img))
else:
cut_imgs.append(self.transforms(cropped))
if block_num == 1:
for eachObject in detections:
cropped = img.crop(eachObject["box_points"])
# print(eachObject["box_points"])
m = eachObject["box_points"][2] - eachObject["box_points"][0]
n = eachObject["box_points"][3] - eachObject["box_points"][1]
# print('截图宽：%d,截图高：%d' % (m, n))
jie = (m * n) / img_area * 100
# print('截图占比：%.2f%%' % jie)
if jie < 10:
cut_imgs.append(self.transforms(img))
else:
cut_imgs.append(self.transforms(cropped))
cut_imgs.append(self.transforms(img))
if block_num == 0:
for i in range(2):
cut_imgs.append(self.transforms(img))
if block_num > 2:
jie_d = {}
for eachObject in detections:
cropped = img.crop(eachObject["box_points"])
# print(eachObject["box_points"])
m = eachObject["box_points"][2] - eachObject["box_points"][0]
n = eachObject["box_points"][3] - eachObject["box_points"][1]
# print('截图宽：%d,截图高：%d' % (m, n))
jie = (m * n) / img_area * 100
# print('截图占比：%.2f%%' % jie)
jie_d[jie] = cropped
jie_l = sorted(jie_d.keys(), reverse=True)
for l in range(2):
if jie_l[l] < 10:
cut_imgs.append(self.transforms(img))
else:
cut_imgs.append(self.transforms(jie_d[jie_l[l]]))
return cut_imgs, index
def __len__(self):
return len(self.imgs)
sigmaM = 0.0001
sigmaZ = 3 * noise
Q = sigmaM**2 * np.eye(4)
R = sigmaZ**2 * np.eye(2)
listCenterX = []
listCenterY = []
listpuntos = []
execution_path = os.getcwd()
detector = ObjectDetection()
detector.setModelTypeAsTinyYOLOv3()
detector.setModelPath(os.path.join(execution_path, "yolo-tiny.h5"))
detector.loadModel(detection_speed="flash")
custom = detector.CustomObjects(sports_ball=True)
cap = cv2.VideoCapture(0)
x = ()
while (True):
ret, frame = cap.read()
img1 = Image.fromarray(frame, 'RGB')
img1.save('image1.jpg')
detections, extracted_objects_array = detector.detectCustomObjectsFromImage(
custom_objects=custom,
input_image=os.path.join(execution_path, "image1.jpg"),
output_image_path=os.path.join(execution_path, "image3new-custom.jpg"),
extract_detected_objects=True,
minimum_percentage_probability=30)
for detection, object_path in zip(detections, extracted_objects_array):
print(object_path)
def car_detections():
images = {}
with open('./detections/jpglist.json', 'r') as f:
jpgs = f.read()
paths = json.loads(jpgs)
imgdet = 0 # downloads es el contador que indica en cual link arranca la descarga de imagenes
images_number = 300000 # aca va el numero de links guardados en el json
for imgdet in range(imgdet, images_number):
# en caso de no poder ver el error en consola y ver en que imagen te quedaste
# creo un archivo que va guardando el ultimo estado de imgdet, o sea la ultima imagen que analicé
images['images'] = imgdet
with open('./detections/info.json', "w") as file:
json.dump(images, file)
realimage = paths[
'jpg' +
str(imgdet
)] # realimage es la imagen original la cual hay que analizar
execution_path = os.getcwd()
detector = ObjectDetection()
detector.setModelTypeAsRetinaNet()
detector.setModelPath(
os.path.join(execution_path, "resnet50_coco_best_v2.0.1.h5"))
detector.loadModel()
print("Analazing", realimage, "/ image", imgdet + 1, "of",
images_number)
# defino de que sitio web proviene para luego colocarlo en el json
if 'ml' in realimage:
website = 'ml'
elif 'olx' in realimage:
website = 'olx'
elif 'demotores' in realimage:
website = 'demotores'
elif 'rosariogarage' in realimage:
website = 'rosariogarage'
else:
website = 'unknown'
custom_objects = detector.CustomObjects(car=True)
detections = detector.detectCustomObjectsFromImage(
input_type="file",
custom_objects=custom_objects,
input_image=os.path.join(execution_path, realimage),
output_image_path=os.path.join("./detections/images/car" +
str(imgdet + 1) + ".jpg"),
output_type="file",
minimum_percentage_probability=90,
extract_detected_objects=False)
if len(detections) == 0: # no detecto autos
os.remove("./detections/images/car" + str(imgdet + 1) +
".jpg") # que borre la foto y continue
print("No cars detected")
continue
else: # si detecto autos
# a continuacion el proceso para obtener el auto de mayor relevancia
wh = []
sub_image = []
for y in range(0, len(detections)):
dots = detections[y]['box_points']
coords = [dots[2] - dots[0], dots[3] - dots[1]]
sub_image += [[coords[0] * coords[1], dots]]
wh += [coords[0] * coords[1]]
maximo = max(wh)
for x in range(0, len(sub_image)):
if maximo == sub_image[x][0]:
maximo = sub_image[x][1]
os.remove("./detections/images/car" + str(imgdet + 1) +
".jpg") # remuevo la imagen con las detecciones
print("Image removed")
shutil.copy(realimage, "./detections/images/car" +
str(imgdet + 1) + ".jpg") # copio la imagen original
print("Original image saved")
# JSON:
# traigo los datos del auto que estan en el .json en la carpeta original para tambien guardarlos en el
# nuevo json
realjson = os.path.dirname(os.path.abspath(
realimage)) + "/meta.json" # obtengo el meta.json que esta en
# en el mismo path que la imagen que estoy analizando
with open(realjson, 'r') as f:
car_data = f.read()
data = json.loads(car_data)
# agrego un par de datos mas al .json
data[
'Web Site'] = website # guardo sitio web de donde proviene la imagen
data[
'boxpoints'] = maximo # guardo los 4 puntos que forman el cuadrilatero donde se encuentra
# el auto principal de la imagen en cuestion
# guardo el .json
with open("./detections/images/car" + str(imgdet + 1) + ".json",
'w') as fp:
json.dump(data, fp)
print("Created .json")
print("End")
class StartWindows(QMainWindow):
def __init__(self, camera=None, parent=None):
super(StartWindows, self).__init__(parent=parent)
self.ui = Ui_Form()
self.ui.setupUi(self)
#detector
#button
self.ui.pushButton.clicked.connect(self.start)
self.ui.pushButton_2.clicked.connect(self.stop)
#camera
self.camera = cv2.VideoCapture(0)
#timer
self.update_timer = QTimer()
self.update_timer.timeout.connect(self.update)
def start(self):
model = self.ui.comboBox.currentText()
print(model)
if model == "YOLO V3":
self.yolo()
elif model == "YOLO TINY":
self.yolo_tiny()
elif model == "RESNET":
self.resnet()
def yolo(self):
self.update_timer.start(30)
self.execution_path = os.getcwd()
self.detector = ObjectDetection()
self.detector.setModelTypeAsYOLOv3()
self.detector.setModelPath(os.path.join(self.execution_path,
"yolo.h5"))
self.detector.loadModel(detection_speed="flash")
print("###you are use yolo model###")
def yolo_tiny(self):
self.update_timer.start(30)
self.execution_path = os.getcwd()
self.detector = ObjectDetection()
self.detector.setModelTypeAsTinyYOLOv3()
self.detector.setModelPath(
os.path.join(self.execution_path, "yolo-tiny.h5"))
self.detector.loadModel(detection_speed="flash")
print("###you are use yolo_tiny model###")
def resnet(self):
self.update_timer.start(30)
self.execution_path = os.getcwd()
self.detector = ObjectDetection()
self.detector.setModelTypeAsRetinaNet()
self.detector.setModelPath(
os.path.join(self.execution_path, "resnet50_coco_best_v2.0.1.h5"))
self.detector.loadModel(detection_speed="fastest")
print("###you are use resnet model###")
def stop(self):
self.update_timer.stop()
def update(self):
ret, frame = self.camera.read()
frame = cv2.flip(frame, 1)
#detected
custom = self.ui.comboBox_2.currentText()
print(custom)
if custom == "Person":
custom_objects = self.detector.CustomObjects(person=True)
elif custom == "orange":
custom_objects = self.detector.CustomObjects(orange=True)
elif custom == "Cell Phone":
custom_objects = self.detector.CustomObjects(cell_phone=True)
detected_image_array, detections = self.detector.detectCustomObjectsFromImage(
custom_objects=custom_objects,
input_type="array",
input_image=frame,
output_type="array")
#detected_image_array, detections = self.detector.detectCustomObjectsFromImage(custom_objects=custom_objects,output_type="array",input_type="array", input_image= frame,display_percentage_probability=True, display_object_name=True)
for eachObject in detections:
print(eachObject["name"], " : ",
eachObject["percentage_probability"], " : ",
eachObject["box_points"])
#resize
detected_image_array = cv2.resize(detected_image_array, (801, 391))
height, width, channel = detected_image_array.shape
bytesPerLine = 3 * width
qImg = QImage(detected_image_array.data, width, height, bytesPerLine,
QImage.Format_RGB888).rgbSwapped()
pixmap01 = QPixmap.fromImage(qImg)
pixmap_image = QPixmap(pixmap01)
self.ui.label.setPixmap(pixmap_image)
self.ui.label.show()
from imageai.Detection import ObjectDetection
import cv2
import requests
import numpy as np
import time
detector = ObjectDetection()
detector.setModelTypeAsYOLOv3()
detector.setModelPath("yolo.h5")
detector.loadModel(detection_speed="normal")
url = ""
t = 0
tim = 0
ti = 0
while t < 10:
#ret, img = cap.read()
img_resp = requests.get(url)
img_arr = np.array(bytearray(img_resp.content), dtype=np.uint8)
img = cv2.imdecode(img_arr, -1)
start = time.clock()
custom = detector.CustomObjects(person=True)
ti = ti + (time.clock() - start)
tim = tim + ti
ti = 0
from imageai.Detection import ObjectDetection
import os
from time import time
execution_path = os.getcwd()
detector = ObjectDetection()
detector.setModelTypeAsRetinaNet()
detector.setModelPath( os.path.join(execution_path , "resnet50_coco_best_v2.0.1.h5"))
detector.loadModel(detection_speed="flash")
our_time = time()
detections = detector.detectObjectsFromImage(input_image=os.path.join(execution_path , "6.jpg"), output_image_path=os.path.join(execution_path , "6flash.jpg"), minimum_percentage_probability=30)
print("IT TOOK : ", time() - our_time)
for eachObject in detections:
print(eachObject["name"] + " : " + eachObject["percentage_probability"] )
print("--------------------------------")
from imageai.Detection import ObjectDetection
import os
import time
dir = ObjectDetection()
dir.setModelTypeAsRetinaNet()
dir.setModelPath(os.path.abspath('retinanet.h5'))
dir.loadModel(detection_speed='faster')
objects = dir.CustomObjects(car=True,
bus=True,
train=True,
truck=True,
traffic_light=True,
stop_sign=True)
print('----- detecting -----')
t = time.time()
det = dir.detectCustomObjectsFromImage(
custom_objects=objects,
input_image=os.path.abspath('imageai_test_2.jpg'),
output_image_path=os.path.abspath('imageai_test_2_res.jpg'))
print(time.time() - t)
for each_obj in det:
print(each_obj)
'''
Editor: Ahan M R
Date: 18-06-2018
Python 3.6.0
'''
#PS1 CROWD DETECTION IN THE SITUATION
'''
From imageAI API, we import Object detection and use it to detect the object classes in the image
'''
from imageai.Detection import ObjectDetection
import os
#os.getcwd gets the current working directory of the image
execution_path = os.getcwd()
detection = ObjectDetection()
detection.setModelTypeAsRetinaNet()
detection.setModelPath( os.path.join(execution_path , "resnet50_coco_best_v2.0.1.h5"))
detection.loadModel()
detections = detection.detectObjectsFromImage(input_image=os.path.join(execution_path , "xi.jpg"), output_image_path=os.path.join(execution_path , "imagenew.jpg"))
for eachObject in detections:
print(eachObject["name"] + " : " + eachObject["percentage_probability"])
detector.loadModel() '''
'''YOLO TINY
model_path="models/yolo-tiny.h5"
detector = ObjectDetection()
detector.setModelTypeAsTinyYOLOv3()
detector.setModelPath(model_path)
detector.loadModel()
'''
'''RESNET'''
model_path="models/resnet50_coco.h5"
detector = ObjectDetection()
detector.setModelTypeAsRetinaNet()
detector.setModelPath(model_path)
detector.loadModel(detection_speed='fast')
video_detector = VideoObjectDetection()
video_detector.setModelTypeAsRetinaNet()
video_detector.setModelPath(model_path)
video_detector.loadModel()
objects_dict = {}
pub_str = rospy.Publisher('objects_detected', String, queue_size=100)
custom = detector.CustomObjects(bottle=True, wine_glass=True,
cup=True, fork=True, knife=True, spoon=True, banana=True, apple=True, sandwich=True, orange=True,
mouse=True, remote=True, cell_phone=True, book=True, scissors=True)
def main():
detector = ObjectDetection()
detector.setModelTypeAsRetinaNet()
detector.setModelPath(
os.path.join(execution_path, "resnet50_coco_best_v2.0.1.h5"))
detector.loadModel("faster")
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
try:
#s.bind((socket.gethostname(), 8000))
s.bind((HOST, 8000))
except socket.error as err:
print("Error de bind")
sys.exit()
s.listen(5)
img = carga_imagen("listo.png")
canvas.create_image(0, 0, image=img, anchor=NW)
text.config(state="normal")
text.delete(1.0, END)
text.insert(INSERT, "Esperando una conexion")
text.config(state="disabled")
while True:
print("esperando una conexion")
clientsocket, address = s.accept()
print(f"Conexion desde {address} establecida ")
f = open("recibido.png", "wb")
img = carga_imagen("cargando1.png")
canvas.create_image(0, 0, image=img, anchor=NW)
full_msg = ''
#msg = clientsocket.recv(1024)
#print(msg)
#temp = msg[22:]
#full_msg += temp.decode("utf-8")
while True:
msg = clientsocket.recv(1024)
#print(msg)
if len(msg) < 1024:
try:
#temp = msg[7:]
#full_msg += temp.decode("utf-8")
full_msg += msg.decode("utf-8")
except:
print("No se pudo decodificar el mensaje")
break
try:
full_msg += msg.decode("utf-8")
except:
print("No se pudo decodificar el mensaje")
#msg = clientsocket.recv(100000)
print(full_msg)
try:
js = literal_eval(str(full_msg))
except:
print("No se pudo convertir el mensaje a formato json")
text.config(state="normal")
text.delete(1.0, END)
text.insert(
INSERT, "Datos sensor ultrasonico: " + str(js["distancia"]) +
"cm de distancia\n\nDatos camara: Fotografia con un peso de " +
str(len(js["imagen"])) + " bytes\n\n")
text.config(state="disabled")
try:
imagen = base64.b64decode(js["imagen"])
f.write(imagen)
detections = detector.detectObjectsFromImage(
input_image=os.path.join(execution_path, "recibido.png"),
output_image_path=os.path.join(execution_path, "imagenew.png"),
minimum_percentage_probability=40)
img = carga_imagen("imagenew.png")
canvas.create_image(0, 0, image=img, anchor=NW)
objetos_detectados = compilar1(detections)
"""if len(detections) != 0:
for eachObject in detections:
objetos_detectados += ""+str(verificarGen(traducir(eachObject["name"])))+ ";" + str(eachObject["percentage_probability"] )+" "
else:
objetos_detectados += "una pared u objeto desconocido "
"""
text.config(state="normal")
text.insert(
INSERT,
"Objeto(s) identificado(s): " + objetos_detectados + "\n\n")
msg_to_send = compilar2(detections, js)
#msg_to_send = direccion(str(js["direccion"]))+"hay "+objetos_detectados+"a "+str(js["distancia"])+"cm de distancia de usted"
print("Natural: " + msg_to_send)
print("\nMorse: " + str(encode(msg_to_send)))
text.insert(
INSERT, "Mensaje en lenguaje natural: " + msg_to_send +
"\n\nMensaje en codigo morse: " + encode(msg_to_send) + "\n\n")
text.config(state="disabled")
clientsocket.send(bytes(msg_to_send, "utf-8"))
clientsocket.close()
except:
print("Error grave")
clientsocket.close()
img = carga_imagen("listo.png")
canvas.create_image(0, 0, image=img, anchor=NW)
text.config(state="normal")
text.delete(1.0, END)
text.insert(INSERT, "Esperando una conexion")
text.config(state="disabled")
#
import speech_recognition as sr
import cv2
import serial
import numpy as np
from imageai.Detection import ObjectDetection
import os
execution_path = os.getcwd()
detector = ObjectDetection()
detector.setModelTypeAsRetinaNet() ## NOTE: Установка модели в качестве.....
detector.setModelPath(
os.path.join(
execution_path,
"C:\\Users\\Admin_Robo\\Desktop\\project\\models\\resnet50_coco_best_v2.0.1.h5"
)) # NOTE: Модели загрузка
detector.loadModel() ## NOTE: + Установка скорости.
print('LOADED')
print('LOADED2')
cap = cv2.VideoCapture()
ret, frame = cap.read()
cv2.imwrite('C:\\Users\\Admin_Robo\\Desktop\\project\\images\\test.png', frame)
cap.release()
cv2.destroyAllWindows()
while True:
## NOTE: Обработка
detections = detector.detectObjectsFromImage(
input_image=os.path.join(
execution_path,
"C:\\Users\\Admin_Robo\\Desktop\\project\\images\\test.png"),
output_image_path=os.path.join(
execution_path,
# ref: https://imageai.readthedocs.io/en/latest/detection/index.html
# This module will count the objects in an image and store them
import time
from imageai.Detection import ObjectDetection
import os
img_detector = ObjectDetection()
img_detector.setModelTypeAsYOLOv3() #147 layers
img_detector.setModelPath(
os.path.join("/catkin_ws/src/master_pkg/src/", "yolo.h5")) #147 layers
img_detector.loadModel(
detection_speed="fast") #normal, fast, faster, fastest, flash
# Note increases in speed should coorespond to lower a 'min. % probability' value
def count(image_array):
try:
det_frame, detections = img_detector.detectObjectsFromImage(
input_type="array",
minimum_percentage_probability=60,
input_image=image_array,
output_type="array")
print("-------------")
for item in detections:
print(item["name"])
print("-------------")
except:
rospy.loginfo("Could not count objects.")
output_file_path= os.path.join(execution_path ,"Detected {}".format(os.path.basename(file_path[:-4]))),
frames_per_second=29, minimum_percentage_probability=min_prob, log_progress=True)
else:
with graph.as_default():
video_path = detector.detectObjectsFromVideo(input_file_path=file_path,
output_file_path=os.path.join(execution_path ,"Detected {}".format(os.path.basename(file_path[:-4]))),
minimum_percentage_probability=min_prob, frames_per_second=29, log_progress=True)
#video_path = detector.detectObjectsFromVideo(input_file_path=file_path,
# output_file_path=os.path.join(execution_path,"Detected {}".format(os.path.basename(file_path[:-4]))),
# frames_per_second=29, log_progress=True)
return(video_path)
detector1 = ObjectDetection()
detector1.setModelTypeAsYOLOv3()
detector1.setModelPath("yolo.h5")
detector1.loadModel()
graph1 = tf.get_default_graph()
result=tuple()
listt=[]
prob={}
def ImageDetection(file_path,objects=None,min_prob=30):
print(file_path)
execution_path = r'./media/output_video'
#detector = ObjectDetection()
#detector.setModelTypeAsYOLOv3()
#detector.setModelPath("yolo.h5")
#detector.loadModel()
if objects:
print(objects)
def func(person,truck,bus,bicycle,bird,motorcycle):
#for k,v in objects.items():"%s = %s" % (k,v)
import matplotlib
matplotlib.use('agg')
import matplotlib.pyplot as plt
from imageai.Detection import ObjectDetection
import os
workdir = os.getcwd()
find_objects = ObjectDetection()
find_objects.setModelTypeAsRetinaNet()
find_objects.setModelPath(os.path.join(workdir,
"resnet50_coco_best_v2.0.1.h5"))
find_objects.loadModel()
objects_found = find_objects.detectObjectsFromImage(
input_image=os.path.join(workdir, "ilia.jpg"),
output_image_path=os.path.join(workdir, "iliaRevealed.jpg"))
for objects in objects_found:
print(objects["name"] + " : " + str(objects["percentage_probability"]))
def getImageDetails():
image = request.files
print()
print(image)
filestr = image['img'].read()
#convert string data to numpy array
npimg = numpy.fromstring(filestr, numpy.uint8)
# convert numpy array to image
print(1)
img = cv2.imdecode(npimg, cv2.IMREAD_UNCHANGED)
print(1)
# img = cv2.imread(filename)
cv2.imshow("image from post", img)
cv2.imwrite("image_from_post_request.jpeg", img)
print(1)
cv2.waitKey(0)
cv2.destroyAllWindows()
# plt.imshow(img)
# plt.show()
filename = "image_from_post_request.jpeg"
execution_path = os.getcwd()
file_name = "image_from_post_request"
detector = ObjectDetection()
detector.setModelTypeAsRetinaNet()
detector.setModelPath(
os.path.join(execution_path, "resnet50_coco_best_v2.0.1.h5"))
detector.loadModel()
detections = detector.detectObjectsFromImage(
input_image=os.path.join(execution_path, file_name + ".jpeg"),
output_image_path=os.path.join(execution_path,
file_name + "detected.jpeg"))
image = cv2.imread(file_name + ".jpeg")
copy = image.copy()
face_cascade = cv2.CascadeClassifier('haarcascade_frontalface_default.xml')
# ROI_number = 0
# return(jsonify({"out" : "Success !!!!"}))
output = {
"humna_face": {
"count": 0
},
"animal": {
"count": 0,
"objects_found": dict()
},
"object": {
"count": 0,
"objects_found": dict()
}
}
animal = [
"bird", "cat", "dog", 'horse', "sheep", "cow", "elephant", "bear",
"zebra", "giraffe"
]
for eachObject in detections:
print(output, eachObject.keys())
print(eachObject["name"], " : ", eachObject["percentage_probability"])
if (eachObject['name'] == 'person'):
x, y, w, h = eachObject['box_points']
ROI = image[y:y + h, x:x + w]
cv2.imwrite('ROI.jpeg', ROI)
cv2.rectangle(copy, (x, y), (x + w, y + h), (36, 255, 12), 2)
person = cv2.imread('ROI.jpeg')
# Convert into grayscale
gray = cv2.cvtColor(person, cv2.COLOR_BGR2GRAY)
# Detect faces
faces = face_cascade.detectMultiScale(gray, 1.4, 1, minSize=(1, 1))
if (len(faces) == 0):
pass
else:
output['human_face']['count'] += 1
elif (eachObject['name'] in animal):
output['animal']['count'] += 1
if (eachObject['name']
not in output['animal']['objects_found'].keys()):
output['animal']['objects_found'][eachObject['name']] = 0
output['animal']['objects_found'][eachObject['name']] += 1
else:
output['object']['count'] += 1
if (eachObject['name']
not in output['object']['objects_found'].keys()):
output['object']['objects_found'][eachObject['name']] = 0
output['object']['objects_found'][eachObject['name']] += 1
# return(send_file(filename_or_fp = file_name + "detected.jpeg", mimetype="image/gif"))
# return(m.to_string(), {'Content-Type': m.content_type})
return (jsonify(output))
# , frames_per_second=29, log_progress=True)
#print(video_path)
from imageai.Detection import ObjectDetection
import os
execution_path = os.getcwd()
detector = ObjectDetection()
detector.setModelTypeAsRetinaNet()
detector.setModelPath(
os.path.join(
execution_path,
"C:\\Users\\Admin_Robo\\Desktop\\resnet50_coco_best_v2.0.1.h5"))
detector.loadModel(
detection_speed='faster'
) ## NOTE: 1 секунда, 5 объектов, самое быстрое для распознавания.
print('LOADED')
print('LOADED2')
detections = detector.detectObjectsFromImage(
input_image=os.path.join(
execution_path,
"C:\\Users\\Admin_Robo\\Desktop\\WIN_20190919_18_18_19_Pro.jpg"),
output_image_path=os.path.join(
execution_path, "C:\\Users\\Admin_Robo\\Desktop\\iimg666.jpg"))
print(
'=================================IMG 1=================================')
for eachObject in detections:
print(eachObject["name"], " : ", eachObject["percentage_probability"])
class disjoin_counter:
def __init__(self, Pass, recording_path):
#intilizing the fundamental veriable
self.recording_path = recording_path
self.Pass = Pass
self.my_col = None
def connect(self):
self.my_col = my_collection(self.Pass)
if self.my_col == -1:
print("Please enter a valid access key !")
exit()
print(" Disjoin counter started successfully!")
def get_detector(self):
# starting ObjectDetection module
self.detector = ObjectDetection()
self.detector.setModelTypeAsRetinaNet()
self.detector.setModelPath('Model/resnet50_coco_best_v2.1.0.h5')
self.detector.loadModel()
def start(self, store_output_video=False):
self.connect()
self.get_detector()
# starting the video capture activity
vid = cv2.VideoCapture('input_video.mp4') #str(self.recording_path))
i = 0
j = 10
while (vid.isOpened()):
ret, frame = vid.read()
if ret == True:
if (j % 10 == 0): # taking every 10th frame
t = datetime.now()
t = t + timedelta(seconds=1)
custom = self.detector.CustomObjects(person=True)
returned_image, detections = self.detector.detectCustomObjectsFromImage(
custom_objects=custom,
input_image=frame,
output_type="array",
minimum_percentage_probability=30,
input_type="array")
out_img = "Store/" + "imagenew" + str(i) + ".jpg"
if store_output_video:
cv2.imwrite(out_img, returned_image)
cv2.imshow('F', returned_image)
i += 1
count = len(detections)
d = t.strftime("%m/%d/%Y, %H:%M:%S")
mydict = {"_id": i, "count": count, "Date_time": d}
x = self.my_col.insert_one(
mydict) # inserting into database
if cv2.waitKey(25) & 0xFF == ord('q'):
break
j += 1
else:
break
vid.release()
#out.release()
cv2.destroyAllWindows()