def test_yolo_infer(self):
from nsrec.nets import simple_yolo
self._create_yolo_test_graph()
inferrable = Inferrable(test_helper.test_graph_file, None,
YOLOToExportModel.INPUT_NODE_NAME, YOLOToExportModel.OUTPUT_NODE_NAME)
input_data = inputs.read_img(os.path.join(test_helper.train_data_dir_path, '1.png'))
input_data = inputs.normalize_img(input_data, [simple_yolo.image_width, simple_yolo.image_height])
pbs = inferrable.infer(np.array([input_data]))
print(pbs.shape)
print(pbs)
def test_infer(self):
from nsrec.nets import lenet_v2
self._create_test_graph()
inferrable = Inferrable(test_helper.test_graph_file,
CNNBBoxToExportModel.INITIALIZER_NODE_NAME,
CNNBBoxToExportModel.INPUT_NODE_NAME, CNNBBoxToExportModel.OUTPUT_NODE_NAME)
input_data = inputs.read_img(os.path.join(test_helper.train_data_dir_path, '1.png'))
input_data = inputs.normalize_img(input_data, [lenet_v2.image_width, lenet_v2.image_height])
pbs = inferrable.infer(np.array([input_data]))
print(pbs)
def test_img_data_generator(new_size, crop_bbox=False):
for i in range(10):
filename = '%s.png' % (i + 1)
img_idx = metadata['filenames'].index(filename)
bbox, label = metadata['bboxes'][img_idx], metadata['labels'][img_idx]
input_data = inputs.read_img(os.path.join(test_helper.train_data_dir_path, filename))
width, height = input_data.shape[1], input_data.shape[0]
if crop_bbox:
input_data = inputs.read_img(os.path.join(test_helper.train_data_dir_path, filename), bbox)
input_data = inputs.normalize_img(input_data, [new_size[0], new_size[1]])
yield (input_data, (width, height), bbox, label)
def infer(self, sess, data):
def to_coordinate_bboxes(label, image_shape):
def to_coordinate_bbox(bbox):
w, h = image_shape[1], image_shape[0]
return list(
map(int,
[bbox[0] * w, bbox[1] * h, bbox[2] * w, bbox[3] * h]))
return [to_coordinate_bbox(l['bbox']) for l in label]
input_data = [
inputs.normalize_img(image, self.config.size) for image in data
]
net_out = sess.run(self.net_out, feed_dict={self.inputs: input_data})
labels = []
for net_out_i in net_out:
labels.append(
extract_label(net_out_i, self.max_number_length,
self.config.num_classes, self.config.threshold))
join_label = lambda label: ''.join(
map(lambda l: str(l['label']), label))
return [(join_label(labels[i]),
to_coordinate_bboxes(labels[i], data[i].shape))
for i in range(len(labels))]