def postprocess(self, boxes, ori_image_shape, threshold):
over_thres_idx = np.nonzero(boxes[:, 1:2] >= threshold)[0]
if len(over_thres_idx) == 0:
pred_dets = np.zeros((1, 6), dtype=np.float32)
pred_xyxys = np.zeros((1, 4), dtype=np.float32)
return pred_dets, pred_xyxys
else:
boxes = boxes[over_thres_idx]
pred_bboxes = boxes[:, 2:]
pred_xyxys, keep_idx = clip_box(pred_bboxes, ori_image_shape)
if len(keep_idx[0]) == 0:
pred_dets = np.zeros((1, 6), dtype=np.float32)
pred_xyxys = np.zeros((1, 4), dtype=np.float32)
return pred_dets, pred_xyxys
pred_scores = boxes[:, 1:2][keep_idx[0]]
pred_cls_ids = boxes[:, 0:1][keep_idx[0]]
pred_tlwhs = np.concatenate(
(pred_xyxys[:, 0:2], pred_xyxys[:, 2:4] - pred_xyxys[:, 0:2] + 1),
axis=1)
pred_dets = np.concatenate((pred_tlwhs, pred_scores, pred_cls_ids),
axis=1)
return pred_dets, pred_xyxys
def postprocess(self,
boxes,
ori_image_shape,
threshold,
inputs,
scaled=False):
over_thres_idx = np.nonzero(boxes[:, 1:2] >= threshold)[0]
if len(over_thres_idx) == 0:
pred_dets = np.zeros((1, 6), dtype=np.float32)
pred_xyxys = np.zeros((1, 4), dtype=np.float32)
return pred_dets, pred_xyxys
else:
boxes = boxes[over_thres_idx]
if not scaled:
# scaled means whether the coords after detector outputs
# have been scaled back to the original image, set True
# in general detector, set False in JDE YOLOv3.
input_shape = inputs['image'].shape[2:]
im_shape = inputs['im_shape'][0]
scale_factor = inputs['scale_factor'][0]
pred_bboxes = scale_coords(boxes[:, 2:], input_shape, im_shape,
scale_factor)
else:
pred_bboxes = boxes[:, 2:]
pred_xyxys, keep_idx = clip_box(pred_bboxes, ori_image_shape)
if len(keep_idx[0]) == 0:
pred_dets = np.zeros((1, 6), dtype=np.float32)
pred_xyxys = np.zeros((1, 4), dtype=np.float32)
return pred_dets, pred_xyxys
pred_scores = boxes[:, 1:2][keep_idx[0]]
pred_cls_ids = boxes[:, 0:1][keep_idx[0]]
pred_tlwhs = np.concatenate(
(pred_xyxys[:, 0:2], pred_xyxys[:, 2:4] - pred_xyxys[:, 0:2] + 1),
axis=1)
pred_dets = np.concatenate((pred_tlwhs, pred_scores, pred_cls_ids),
axis=1)
return pred_dets, pred_xyxys
def reidprocess(self, det_results, repeats=1):
pred_dets = det_results['boxes'] # cls_id, score, x0, y0, x1, y1
pred_xyxys = pred_dets[:, 2:6]
ori_image = det_results['ori_image']
ori_image_shape = ori_image.shape[:2]
pred_xyxys, keep_idx = clip_box(pred_xyxys, ori_image_shape)
if len(keep_idx[0]) == 0:
det_results['boxes'] = np.zeros((1, 6), dtype=np.float32)
det_results['embeddings'] = None
return det_results
pred_dets = pred_dets[keep_idx[0]]
pred_xyxys = pred_dets[:, 2:6]
w, h = self.tracker.input_size
crops = get_crops(pred_xyxys, ori_image, w, h)
# to keep fast speed, only use topk crops
crops = crops[:50] # reid_batch_size
det_results['crops'] = np.array(crops).astype('float32')
det_results['boxes'] = pred_dets[:50]
input_names = self.reid_predictor.get_input_names()
for i in range(len(input_names)):
input_tensor = self.reid_predictor.get_input_handle(input_names[i])
input_tensor.copy_from_cpu(det_results[input_names[i]])
# model prediction
for i in range(repeats):
self.reid_predictor.run()
output_names = self.reid_predictor.get_output_names()
feature_tensor = self.reid_predictor.get_output_handle(
output_names[0])
pred_embs = feature_tensor.copy_to_cpu()
det_results['embeddings'] = pred_embs
return det_results