class YoloObjectDetector:
def __init__(self, modelSize):
if not (os.path.exists("models/yolo-tiny/yolov3-tiny.weights")
and os.path.exists("models/yolo/yolov3.weights")):
sys.exit(
"Please download the YOLOv3 weight files first (using downloadYOLO.sh)"
)
if modelSize == "tiny":
self.modelSize = 416
self.model = cv2.dnn.readNet(
"models/yolo-tiny/yolov3-tiny.weights",
"models/yolo-tiny/yolov3-tiny.cfg")
else:
self.modelSize = int(modelSize)
self.model = cv2.dnn.readNet("models/yolo/yolov3.weights",
"models/yolo/yolov3.cfg")
with open("models/yolo/yolov3.txt", 'r') as f:
self.classes = [line.strip() for line in f.readlines()]
self.output = Output(self.classes)
# scale is 0.00392 for YOLO as it does not use 0..255 but 0..1 as range (0.00392 = 1/255)
self.scale = 0.00392
def processImage(self, image):
if image is None:
print("Ignoring image")
return
image_height, image_width, _ = image.shape
blob = cv2.dnn.blobFromImage(image,
self.scale,
(self.modelSize, self.modelSize),
(0, 0, 0),
True,
crop=False)
self.model.setInput(blob)
retval = self.model.forward(self.get_output_layers(self.model))
class_ids = []
confidences = []
boxes = []
conf_threshold = 0.5
nms_threshold = 0.4
for out in retval:
for detection in out:
scores = detection[5:]
class_id = np.argmax(scores)
confidence = scores[class_id]
if confidence > conf_threshold:
center_x = int(detection[0] * image_width)
center_y = int(detection[1] * image_height)
w = int(detection[2] * image_width)
h = int(detection[3] * image_height)
x = center_x - w / 2
y = center_y - h / 2
class_ids.append(class_id)
confidences.append(float(confidence))
boxes.append([x, y, w, h])
indices = cv2.dnn.NMSBoxes(boxes, confidences, conf_threshold,
nms_threshold)
if len(indices) == 0:
return
for i in indices:
i = i[0]
box = boxes[i]
x = box[0]
y = box[1]
w = box[2]
h = box[3]
self.output.draw_prediction(image, class_ids[i], confidences[i],
round(x), round(y), round(x + w),
round(y + h))
return image
def get_output_layers(self, net):
layer_names = net.getLayerNames()
output_layers = [
layer_names[i[0] - 1] for i in net.getUnconnectedOutLayers()
]
return output_layers
