def local_eval(func, model, image_size, test_path, name_path, verbose):
tmp_path = os.path.join('tmp' +
time.strftime("%Y%m%d%H%M", time.localtime()))
if os.path.exists(tmp_path):
os.remove(tmp_path)
with open(test_path) as f:
lines = f.readlines()
paths = [line.split()[0] for line in lines]
infer_time = []
with open(tmp_path, 'a+') as f:
for i, path in enumerate(paths, 1):
if i == 1:
sys.stdout.write('\n')
sys.stdout.write('\r' + keras_bar(i, len(paths)))
image = read_image(path)
h, w = image.shape[:2]
image = preprocess_image(image, (image_size, image_size)).astype(
np.float32)
images = np.expand_dims(image, axis=0)
tic = time.time()
bboxes, scores, classes, valid_detections = model.predict(images)
toc = time.time()
infer_time.append(toc - tic)
bboxes = bboxes[0][:valid_detections[0]]
scores = scores[0][:valid_detections[0]]
classes = classes[0][:valid_detections[0]]
bboxes *= image_size
_, bboxes = preprocess_image_inv(image, (w, h), bboxes)
line = path
for bbox, score, cls in zip(bboxes, scores, classes):
x1, y1, x2, y2 = bbox
line += " {:.2f},{:.2f},{:.2f},{:.2f},{},{:.4f}".format(
x1, y1, x2, y2, int(cls), score)
f.write(line + '\n')
ans = func(test_path, tmp_path, name_path, verbose)
# remove tmp
os.remove(tmp_path)
if verbose:
if len(infer_time) > 5:
s = np.mean(infer_time[5:])
else:
s = np.mean(infer_time)
print('Inference time', s * 1000, 'ms')
return ans
def _getitem(self, sub_idx):
path, bboxes, labels = self.annotation[sub_idx]
image = read_image(path)
if len(bboxes) != 0:
bboxes, labels = np.array(bboxes), np.array(labels)
else:
bboxes, labels = np.zeros((0, 4)), np.zeros((0,))
image, bboxes = preprocess_image(image, (self._image_size, self._image_size), bboxes)
labels = augment.onehot(labels, self.num_classes, self.label_smoothing)
return image, bboxes, labels
def inference(image):
h, w = image.shape[:2]
image = preprocess_image(image, (image_size, image_size)).astype(np.float32)
images = np.expand_dims(image, axis=0)
tic = time.time()
bboxes, scores, classes, valid_detections = model.predict(images)
toc = time.time()
bboxes = bboxes[0][:valid_detections[0]]
scores = scores[0][:valid_detections[0]]
classes = classes[0][:valid_detections[0]]
# bboxes *= image_size
_, bboxes = postprocess_image(image, (w, h), bboxes)
return (toc - tic) * 1000, bboxes, scores, classes
def _getitem(self, sub_idx):
path, bboxes, labels = self.annotation[sub_idx]
image = read_image(path)
bboxes, labels = np.array(bboxes), np.array(labels)
if self.normal_method:
image = augment.random_distort(image)
image = augment.random_grayscale(image)
image, bboxes = augment.random_flip_lr(image, bboxes)
image, bboxes = augment.random_rotate(image, bboxes)
image, bboxes, labels = augment.random_crop_and_zoom(
image, bboxes, labels, (self._image_size, self._image_size))
else:
image, bboxes = preprocess_image(
image, (self._image_size, self._image_size), bboxes)
labels = augment.onehot(labels, self.num_classes, self.label_smoothing)
return image, bboxes, labels