def generate_batch(image):
"""
preprocess image, return batch
:param im: cv2.imread returns [height, width, channel] in BGR
:return:
data_batch: MXNet input batch
data_names: names in data_batch
im_scale: float number
"""
data = []
data_shapes = []
im_scale = []
data_name = []
t1 = []
t2 = []
for i in range(0, image.qsize()):
im = image.get()
im_array, im_scalet = resize(im, SHORT_SIDE, LONG_SIDE)
im_array = transform(im_array, PIXEL_MEANS)
im_info = np.array([[im_array.shape[2], im_array.shape[3], im_scalet]],
dtype=np.float32)
t1.append(im_array.shape)
t2.append(im_info.shape)
data.append([mx.nd.array(im_array), mx.nd.array(im_info)])
data_shapes.append([('data', im_array.shape),
('im_info', im_info.shape)])
im_scale.append(im_scalet)
data_name.append(DATA_NAMES)
data_batch = mx.io.DataBatch(data=data,
label=None,
provide_data=data_shapes,
provide_label=None)
return data_batch, data_name, im_scale
def generate_batch(im):
"""
preprocess image, return batch
:param im: cv2.imread returns [height, width, channel] in BGR
:return:
data_batch: MXNet input batch
data_names: names in data_batch
im_scale: float number
"""
im_array, im_scale = resize(im, SHORT_SIDE, LONG_SIDE)
im_info = np.array([[im_array.shape[0], im_array.shape[1], im_scale]], dtype=np.float32)
if im_array.shape[0] == SHORT_SIDE or im_array.shape[0] == LONG_SIDE:
if im_array.shape[1] % 32 != 0:
pad_size = int(math.ceil(im_array.shape[1] / 32.) * 32 - im_array.shape[1])
im_array = cv2.copyMakeBorder(im_array, 0, 0, 0, pad_size, cv2.BORDER_CONSTANT, value=(0, 0, 0))
elif im_array.shape[1] == SHORT_SIDE or im_array.shape[1] == LONG_SIDE:
if im_array.shape[0] % 32 != 0:
pad_size = int(math.ceil(im_array.shape[0] / 32.) * 32 - im_array.shape[0])
im_array = cv2.copyMakeBorder(im_array, 0, pad_size, 0, 0, cv2.BORDER_CONSTANT, value=(0, 0, 0))
im_array = transform(im_array, config.PIXEL_MEANS)
# im_info = np.array([[im_array.shape[2], im_array.shape[3], im_scale]], dtype=np.float32)
data = [mx.nd.array(im_array), mx.nd.array(im_info)]
data_shapes = [('data', im_array.shape), ('im_info', im_info.shape)]
data_batch = mx.io.DataBatch(data=data, label=None, provide_data=data_shapes, provide_label=None)
return data_batch, DATA_NAMES, im_scale
def generate_batch(im):
"""
preprocess image, return batch
:param im: cv2.imread returns [height, width, channel] in BGR
:return:
data_batch: MXNet input batch
data_names: names in data_batch
im_scale: float number
"""
import ipdb
ipdb.set_trace()
im_array, im_scale = resize(im,
SHORT_SIDE,
LONG_SIDE,
stride=config.IMAGE_STRIDE)
im_array = transform(im_array, PIXEL_MEANS)
im_info = np.array([[im_array.shape[2], im_array.shape[3], im_scale]],
dtype=np.float32)
data = [mx.nd.array(im_array), mx.nd.array(im_info)]
data_shapes = [('data', im_array.shape), ('im_info', im_info.shape)]
data_batch = mx.io.DataBatch(data=data,
label=None,
provide_data=data_shapes,
provide_label=None)
return data_batch, DATA_NAMES, im_scale
def generate_batch(im):
"""
preprocess image, return batch
:param im: cv2.imread returns [height, width, channel] in BGR
:return:
data_batch: MXNet input batch
data_names: names in data_batch
im_scale: float number
"""
im_array, im_scale = resize(im, SHORT_SIDE, LONG_SIDE)
im_array = transform(im_array, PIXEL_MEANS)
im_info = np.array([[im_array.shape[2], im_array.shape[3], im_scale]], dtype=np.float32)
data = [mx.nd.array(im_array), mx.nd.array(im_info)]
data_shapes = [('data', im_array.shape), ('im_info', im_info.shape)]
data_batch = mx.io.DataBatch(data=data, label=None, provide_data=data_shapes, provide_label=None)
return data_batch, DATA_NAMES, im_scale
def load_data():
# load demo data
demo_folder = '../data/demo/'
image_list = os.listdir(demo_folder)
data = []
im_scales = []
for im_name in image_list:
assert os.path.exists(demo_folder + im_name), ('%s does not exist'.format(demo_folder + im_name))
im = cv2.imread(demo_folder + im_name, cv2.IMREAD_COLOR) # | cv2.IMREAD_IGNORE_ORIENTATION)
target_size = config.SCALES[0][0]
max_size = config.SCALES[0][1]
im, im_scale = resize(im, target_size, max_size, stride=config.IMAGE_STRIDE)
im_tensor = transform(im, config.PIXEL_MEANS)
im_info = np.array([[im_tensor.shape[2], im_tensor.shape[3], im_scale]], dtype=np.float32)
data.append({'data': im_tensor, 'im_info': im_info})
im_scales.append(im_scale)
return data, image_list, im_scales
