class ParkingSlotDetection:
def __init__(self, labels_path=labels_camera1_path, thresh=0.31, rois=def_rois,v1=False,camera_id=None):
self.camera_id = camera_id
self.thresh = thresh
self.count = 0
execution_path = os.getcwd()
self.v1 = v1
if v1:
self.detector = ObjectDetection()
self.detector.setModelTypeAsRetinaNet()
self.detector.setModelPath(os.path.join(execution_path, "resnet50_coco_best_v2.1.0.h5"))
self.detector.loadModel()
self.custom_objects = self.detector.CustomObjects(car=True, bus=True, truck=True)
else:
IMAGE_WIDTH = 110
IMAGE_HEIGHT = 110
self.dim =(IMAGE_WIDTH, IMAGE_HEIGHT)
self.detector = load_model(os.path.join(execution_path,'model_ivision_4.h5'))
self.load_labels(labels_path)
self.rois = rois
# self.parking_places = []
# for file in os.listdir(labels_path):
# if file.endswith(".xml"):
# self.parking_places.append(get_parking_boxes(os.path.join(labels_path, file)))
# # print(file," ",len(parking_places[len(parking_places)-1]))
# self.fill_percentage = []
# temp = []
# for i in range(len(self.parking_places[0])):
# temp.append(0.0)
# for i in range(len(self.parking_places)):
# self.fill_percentage.append(copy.deepcopy(temp))
def load_labels(self, labels_path):
self.parking_places = []
for file in os.listdir(labels_path):
if file.endswith(".xml"):
self.parking_places.append(get_parking_boxes(os.path.join(labels_path, file)))
print(file," ",len(self.parking_places[len(self.parking_places)-1]))
self.fill_percentage = []
temp = []
max_val = max(len(i) for i in self.parking_places)
#print(max_val)
for i in range(max_val):
temp.append(0.0)
for i in range(len(self.parking_places)):
self.fill_percentage.append(copy.deepcopy(temp))
def load_rois(self, rois):
self.rois = rois
def processing(self, image):
reset(self.fill_percentage)
for (x, y, w, h) in self.rois:
roi = image[y:y + h, x:x + w]
image_copy, detections = self.detector.detectObjectsFromImage(custom_objects=self.custom_objects,
input_type="array",
input_image=roi,
output_type="array",
minimum_percentage_probability=30)
for eachObject in detections:
box = eachObject["box_points"]
box[0] += x
box[1] += y
box[2] += x
box[3] += y
# x0, y0, x1, y1 = box
# color = (255, 255, 10)
# thickness = 2
# image = cv2.rectangle(image, (x0, y0), (x1, y1), color, thickness)
for i in range(len(self.parking_places)):
#print("len:",len(self.parking_places[i]))
for j in range(len(self.parking_places[i])):
#print(i," ",j)
self.fill_percentage[i][j] += bb_intersection_over_union(box, self.parking_places[i][j])
# for i in self.fill_percentage:
# print(i)
counts = []
for i in range(len(self.fill_percentage)):
count = 0
for index,j in enumerate(self.fill_percentage[i]):
if index < len(self.parking_places[i]):
if j < self.thresh: count += 1
counts.append(count)
copy_img = image.copy()
ind = np.argmin(counts)
show_value = counts[ind]
show_value = str(show_value) + " (min)"
min_image = draw_result(image, show_value, ind, fill_percentage=self.fill_percentage, thresh=self.thresh,
parking_places=self.parking_places)
print("min:", self.fill_percentage[ind])
ind = np.argmax(counts)
show_value = counts[ind]
show_value = str(show_value) + " (max)"
max_image = draw_result(copy_img, show_value, ind, fill_percentage=self.fill_percentage, thresh=self.thresh,
parking_places=self.parking_places)
print("max:", self.fill_percentage[ind])
numpy_horizontal_concat = np.concatenate((min_image, max_image), axis=0)
return numpy_horizontal_concat
def process_v2(self,image):
self.count+=1
t = time()
pred = []
for part in self.parking_places:
batch = []
for place in part:
x, y, w, h = place
w = w - x
h = h - y
roi = image[y:y + h, x:x + w]
roi = cv2.resize(roi, self.dim, interpolation=cv2.INTER_CUBIC)
batch.append(roi)
batch = np.stack(batch)
# print(len(batch))
y_pred = self.detector.predict(batch, batch_size=len(batch), verbose=0)
y_pred_bool = np.argmax(y_pred, axis=1)
pred.append(y_pred_bool)
s = []
for i in pred:
s.append(np.sum(i))
max_image = image.copy()
min_image = image.copy()
ind = np.argmax(s)
show_value = s[ind]
show_value = str(show_value) + " (max)"
max_image = draw_result_v2(max_image,ind,pred,self.parking_places,show_value)
max_image = cv2.resize(max_image,(1080,720))
if(self.count%30==0):
cv2.imwrite(os.path.join(os.getcwd(), "image_" + str(self.camera_id) + ".jpg"), max_image)
with open(os.path.join(os.getcwd(),'data.json'), 'w') as f:
json.dump({"value":str(s[ind])}, f)
ind = np.argmin(s)
show_value = s[ind]
show_value = str(show_value) + " (min)"
min_image = draw_result_v2(min_image,ind,pred,self.parking_places,show_value)
min_image = cv2.resize(min_image, (1080, 720))
numpy_horizontal_concat = np.concatenate((min_image, max_image), axis=0)
print(1./(time()-t))
# time.sleep(0.1)
return numpy_horizontal_concat
