def __call__(self, img):
if img is None:
return None
(height, width, channel) = img.shape
preds = self.predict(img)
detections = []
for *xyxy, conf, cls in reversed(preds):
xyxy[0] = int(xyxy[0].to('cpu').detach().numpy().copy())
xyxy[1] = int(xyxy[1].to('cpu').detach().numpy().copy())
xyxy[2] = int(xyxy[2].to('cpu').detach().numpy().copy())
xyxy[3] = int(xyxy[3].to('cpu').detach().numpy().copy())
conf = conf.to('cpu').detach().numpy()
label = f'{self.names[int(cls)]}'
list = {
'bbox': xyxy,
'rbbox': bbox_to_rbbox(xyxy, img),
'label': int(cls),
'label_name': f'{self.names[int(cls)]}',
'prob': conf,
'color': self.colors[int(cls)]
}
detections.append(list)
#print('detections :{}'.format(detections))
return [detections]
def update_target_detection_with_tracking(self):
if self.frame_id % 1 == 0 or not self.is_tracking:
self.is_tracking = None
self.target_detection = self._object_detection()
print('target detections :{}'.format(self.target_detection))
if self.target_detection is not None:
self.tracker = cv2.TrackerKCF_create()
self.tracker.init(self.cv_image,
bbox_to_roi(self.target_detection['bbox']))
if self.target_detection is not None:
self.is_tracking, roi = self.tracker.update(self.cv_image)
self.target_detection['bbox'] = roi_to_bbox(roi)
self.target_detection['rbbox'] = bbox_to_rbbox(
self.target_detection['bbox'], self.cv_image)
def update_target_detection(self):
self.target_detection = self._object_detection()
print('target detections :{}'.format(self.target_detection))
if self.target_detection is not None:
self.target_detection['rbbox'] = bbox_to_rbbox(
self.target_detection['bbox'], self.cv_image)