def decode_netout(self, netout, obj_threshold=0.3, nms_threshold=0.3):
grid_h, grid_w, nb_box = netout.shape[:3]
boxes = []
# decode the output by the network
netout[..., 4] = self.sigmoid(netout[..., 4])
netout[..., 5:] = netout[..., 4][..., np.newaxis] * self.softmax(
netout[..., 5:])
netout[..., 5:] *= netout[..., 5:] > obj_threshold
for row in range(grid_h):
for col in range(grid_w):
for b in range(nb_box):
# from 4th element onwards are confidence and class classes
classes = netout[row, col, b, 5:]
if np.sum(classes) > 0:
# first 4 elements are x, y, w, and h
x, y, w, h = netout[row, col, b, :4]
x = (col + self.sigmoid(x)
) / grid_w # center position, unit: image width
y = (row + self.sigmoid(y)
) / grid_h # center position, unit: image height
w = self.anchors[2 * b + 0] * np.exp(
w) / grid_w # unit: image width
h = self.anchors[2 * b + 1] * np.exp(
h) / grid_h # unit: image height
confidence = netout[row, col, b, 4]
box = BoundBox(x, y, w, h, confidence, classes)
boxes.append(box)
# suppress non-maximal boxes
for c in range(self.nb_class):
sorted_indices = list(
reversed(np.argsort([box.classes[c] for box in boxes])))
for i in range(len(sorted_indices)):
index_i = sorted_indices[i]
if boxes[index_i].classes[c] == 0:
continue
else:
for j in range(i + 1, len(sorted_indices)):
index_j = sorted_indices[j]
if self.bbox_iou(boxes[index_i],
boxes[index_j]) >= nms_threshold:
boxes[index_j].classes[c] = 0
# remove the boxes which are less likely than a obj_threshold
boxes = [box for box in boxes if box.get_score() > obj_threshold]
return boxes
def decode_netout(self, netout, obj_threshold=0.3, nms_threshold=0.3):
grid_h, grid_w, nb_box = netout.shape[:3]
boxes = []
# decode the output by the network
netout[..., 4] = self.sigmoid(netout[..., 4])
netout[..., 5:] = netout[..., 4][..., np.newaxis] * self.softmax(netout[..., 5:])
netout[..., 5:] *= netout[..., 5:] > obj_threshold
for row in range(grid_h):
for col in range(grid_w):
for b in range(nb_box):
# from 4th element onwards are confidence and class classes
classes = netout[row,col,b,5:]
if np.sum(classes) > 0:
# first 4 elements are x, y, w, and h
x, y, w, h = netout[row,col,b,:4]
x = (col + self.sigmoid(x)) / grid_w # center position, unit: image width
y = (row + self.sigmoid(y)) / grid_h # center position, unit: image height
w = self.anchors[2 * b + 0] * np.exp(w) / grid_w # unit: image width
h = self.anchors[2 * b + 1] * np.exp(h) / grid_h # unit: image height
confidence = netout[row,col,b,4]
box = BoundBox(x, y, w, h, confidence, classes)
boxes.append(box)
# suppress non-maximal boxes
for c in range(self.nb_class):
sorted_indices = list(reversed(np.argsort([box.classes[c] for box in boxes])))
for i in xrange(len(sorted_indices)):
index_i = sorted_indices[i]
if boxes[index_i].classes[c] == 0:
continue
else:
for j in xrange(i+1, len(sorted_indices)):
index_j = sorted_indices[j]
if self.bbox_iou(boxes[index_i], boxes[index_j]) >= nms_threshold:
boxes[index_j].classes[c] = 0
# remove the boxes which are less likely than a obj_threshold
boxes = [box for box in boxes if box.get_score() > obj_threshold]
return boxes
def decode_netout(self,
netout,
obj_threshold=0.25,
nms_threshold=0.7,
hand_label=-1):
grid_h, grid_w, nb_box = netout.shape[:3]
boxes = []
# decode the output by the network
netout[..., 4] = self.sigmoid(netout[..., 4])
netout[..., 5:] = netout[..., 4][..., np.newaxis] * self.softmax(
netout[..., 5:])
netout[..., 5:] *= netout[..., 5:] > obj_threshold
for row in range(grid_h):
for col in range(grid_w):
for b in range(nb_box):
# from 4th element onwards are confidence and class classes
classes = netout[row, col, b, 5:]
if np.sum(classes) > 0:
# first 4 elements are x, y, w, and h
x, y, w, h = netout[row, col, b, :4]
x = (col + self.sigmoid(x)
) / grid_w # center position, unit: image width
y = (row + self.sigmoid(y)
) / grid_h # center position, unit: image height
w = self.anchors[2 * b + 0] * np.exp(
w) / grid_w # unit: image width
h = self.anchors[2 * b + 1] * np.exp(
h) / grid_h # unit: image height
confidence = netout[row, col, b, 4]
box = BoundBox(x, y, w, h, confidence, classes)
boxes.append(box)
# suppress non-maximal boxes
for c in range(self.nb_class):
sorted_indices = list(
reversed(np.argsort([box.classes[c] for box in boxes])))
for i in range(len(sorted_indices)):
index_i = sorted_indices[i]
if boxes[index_i].classes[c] == 0:
continue
else:
for j in range(i + 1, len(sorted_indices)):
index_j = sorted_indices[j]
if self.bbox_iou(boxes[index_i],
boxes[index_j]) >= nms_threshold:
boxes[index_j].classes[c] = 0
# remove the boxes which are less likely than a obj_threshold
boxes = [box for box in boxes if box.get_score() > obj_threshold]
if len(boxes) == 0:
return boxes
# select highest score of left and right
filtered_boxes = []
max_score_r = 0
max_score_l = 0
count_l_index = -1
count_r_index = -1
for i in range(len(boxes)):
if boxes[i].get_label() == 0:
if boxes[i].get_score() > max_score_l:
max_score_l = boxes[i].get_score()
count_l_index = i
elif boxes[i].get_label() == 1:
if boxes[i].get_score() > max_score_r:
max_score_r = boxes[i].get_score()
count_r_index = i
else:
print('label: ', boxes[i].get_label())
if hand_label == 0:
if count_l_index < 0:
# Ugly
box = boxes[count_r_index]
box.label = 0
filtered_boxes.append(box)
else:
filtered_boxes = [boxes[count_l_index]]
elif hand_label == 1:
if count_r_index < 0:
# Ugly
box = boxes[count_l_index]
box.label = 1
filtered_boxes.append(box)
else:
filtered_boxes = [boxes[count_r_index]]
elif hand_label == 2:
if count_l_index < 0:
print('error bl')
if count_r_index < 0:
print('error br')
#if count_l_index >= 0:
filtered_boxes.append(boxes[count_l_index])
#if count_r_index >= 0:
filtered_boxes.append(boxes[count_r_index])
#if bbox_iou(boxes[count_l_index] != boxes[count_r_index]) < 0.8:
# boxes = [boxes[count_l_index], boxes[count_r_index]]
#else:
elif hand_label == -1:
filtered_boxes.append(boxes[count_l_index])
filtered_boxes.append(boxes[count_r_index])
return filtered_boxes