def __init__(self, config, connector_type):
super().__init__()
self.__config = config
self.input_size = 416
onnx_fn = "/data/yolov4_e_scooter.onnx"
## cpu'da daha hızlı ama false positive fazla
# self.input_size = 640
# onnx_fn = "/data/yolov5_e_scooter.onnx"
onnx_fn = os.path.dirname(os.path.abspath(__file__)) + onnx_fn.replace(
"/", os.path.sep)
classes_fn = "/data/class.names"
classes_fn = os.path.dirname(
os.path.abspath(__file__)) + classes_fn.replace("/", os.path.sep)
self.class_names = onnx_helper.parse_class_names(classes_fn)
self.sess_tuple = onnx_helper.get_sess_tuple(onnx_fn)
# region Demo
self._last_sent = None
self._last_candidate = None
self._last_candidate_counter = 0
def __init__(self, config, _connector_type):
super().__init__()
# # onnx_fn = "/data/small_30.onnx"
# onnx_fn = "/data/best.onnx"
# # self. = "/data/last.onnx"
# self.input_size = 640
onnx_fn = "/data/yolov4_helmet.onnx"
self.input_size = 416
self.average_of_frames = config.get("average_of_frames", 1)
self.confirm_count = config.get("confirm_count", 1)
self.confirm_val = 0
self._use_person = config.get("use_person", True)
if not os.path.isfile(onnx_fn):
onnx_fn = os.path.dirname(
os.path.abspath(__file__)) + onnx_fn.replace("/", os.path.sep)
classes_filename = "/data/class.names"
if not os.path.isfile(classes_filename):
classes_filename = os.path.dirname(
os.path.abspath(__file__)) + classes_filename.replace(
"/", os.path.sep)
self.class_names = onnx_helper.parse_class_names(classes_filename)
self.sess_tuple = onnx_helper.get_sess_tuple(onnx_fn)
self.debug_mode = NDUUtility.is_debug_mode()
self.last_data = None
self.queue = []
def __init__(self, config, connector_type):
super().__init__()
self.__config = config
self.input_size = config.get("input_size", 512)
#### OLD
# # 416 msec üst üste var
# self.input_size = 512
# onnx_fn = "/data/yolov4_-1_3_512_512_dynamic.onnx"
# # 550 msec başarılı (default)
# self.input_size = 608
# onnx_fn = "/data/yolov4_-1_3_608_608_dynamic.onnx"
# ### başarısız
# # 335 msec
# self.input_size = 512
# onnx_fn = "/data/old/yolov4-csp_1_3_512_512_static.onnx"
# # 793 msec üst üste çok var
# self.input_size = 640
# onnx_fn = "/data/yolov4x-mish_1_3_640_640_static.onnx"
# ### OK
# 550 msec başarılı - vino onnx 350
self.input_size = 608
onnx_fn = "/data/yolov4-1_3_608_608_static.onnx"
# # 22 msec güzel - vino onnx 16
# self.input_size = 416
# onnx_fn = "/data/yolov4-tiny_1_3_416_416_static.onnx"
# # vehicles
# self.input_size = 416
# onnx_fn = "/data/yolov4-tiny_vehicles_416_static.onnx"
# # vehicles
# self.input_size = 416
# onnx_fn = "/data/yolov4-tiny_vehicle_lp_7_416_static.onnx"
if not os.path.isfile(onnx_fn):
onnx_fn = os.path.dirname(os.path.abspath(__file__)) + onnx_fn.replace("/", os.path.sep)
classes_filename = config.get("classes_filename", "/data/coco.names")
if not os.path.isfile(classes_filename):
classes_filename = os.path.dirname(os.path.abspath(__file__)) + classes_filename.replace("/", os.path.sep)
self.class_names = onnx_helper.parse_class_names(classes_filename)
self.sess_tuple = onnx_helper.get_sess_tuple(onnx_fn)
def __init__(self, config, connector_type):
super().__init__()
onnx_fn = "/data/yolov5s.onnx"
classes_filename = "/data/coco.names"
self.input_size = 640
if not os.path.isfile(onnx_fn):
onnx_fn = os.path.dirname(os.path.abspath(__file__)) + onnx_fn.replace("/", os.path.sep)
if not os.path.isfile(classes_filename):
classes_filename = os.path.dirname(os.path.abspath(__file__)) + classes_filename.replace("/", os.path.sep)
self.class_names = onnx_helper.parse_class_names(classes_filename)
self.sess_tuple = onnx_helper.get_sess_tuple(onnx_fn)
def __init__(self, config, _connector_type):
super().__init__()
onnx_fn = "/data/yolov4_aerial_vehicle.onnx"
self.input_size = 416
if not os.path.isfile(onnx_fn):
onnx_fn = os.path.dirname(os.path.abspath(__file__)) + onnx_fn.replace("/", os.path.sep)
classes_filename = "/data/class.names"
if not os.path.isfile(classes_filename):
classes_filename = os.path.dirname(os.path.abspath(__file__)) + classes_filename.replace("/", os.path.sep)
self.class_names = onnx_helper.parse_class_names(classes_filename)
self.sess_tuple = onnx_helper.get_sess_tuple(onnx_fn)
def __init__(self, config, connector_type):
super().__init__()
self.__config = config
self.input_size = config.get("input_size", 416)
onnx_fn = config.get("onnx_fn", "/data/yolov3.onnx")
if not os.path.isfile(onnx_fn):
onnx_fn = os.path.dirname(os.path.abspath(__file__)) + onnx_fn.replace("/", os.path.sep)
classes_filename = config.get("classes_filename", "coco.names")
if not os.path.isfile(classes_filename):
classes_filename = os.path.dirname(os.path.abspath(__file__)) + classes_filename.replace("/", os.path.sep)
self.class_names = onnx_helper.parse_class_names(classes_filename)
self.sess_tuple = onnx_helper.get_sess_tuple(onnx_fn)
def __init__(self, config, connector_type):
super().__init__()
self.__config = config
self.__connector_type = connector_type
onnx_fn = path.dirname(path.abspath(__file__)) + "/data/emotion-ferplus-8.onnx"
class_names_fn = path.dirname(path.abspath(__file__)) + "/data/emotion.names"
if not path.isfile(onnx_fn):
raise FileNotFoundError(errno.ENOENT, os.strerror(errno.ENOENT), onnx_fn)
if not path.isfile(class_names_fn):
raise FileNotFoundError(errno.ENOENT, os.strerror(errno.ENOENT), class_names_fn)
self.class_names = onnx_helper.parse_class_names(class_names_fn)
self.sess_tuple = onnx_helper.get_sess_tuple(onnx_fn)
def __init__(self, config, connector_type):
super().__init__()
# onnx_fn = "/data/transpalet_last.onnx"
onnx_fn = "/data/transpalet_best.onnx"
self.input_size = 608
if not os.path.isfile(onnx_fn):
onnx_fn = os.path.dirname(
os.path.abspath(__file__)) + onnx_fn.replace("/", os.path.sep)
classes_filename = "/data/class.names"
if not os.path.isfile(classes_filename):
classes_filename = os.path.dirname(
os.path.abspath(__file__)) + classes_filename.replace(
"/", os.path.sep)
self.class_names = onnx_helper.parse_class_names(classes_filename)
self.sess_tuple = onnx_helper.get_sess_tuple(
onnx_fn, config.get("max_engine_count", 0))
def __init__(self, config, connector_type):
super().__init__()
self.__connector_type = connector_type
self.__dont_use_face_rects = config.get("dont_use_face_rects", False)
self.__upscale_person = config.get("upscale_person", False)
self.__upscale_person = True ####################
self._debug = True #####
self._max_rect = config.get(
"max_rect", 0
) # örneğin 0.5 verilirse, max dim değeri frame'in yarısından büyük olan bbox'lar kabul edilmez.
self._last_data = None
self.confirm_count = config.get("confirm_count", 1)
self.confirm_val = 0
self.time_start = 0
self.start_cooldown_next_iter = False
self.cooldown_timer = 0
self.original_frame = ''
self.capture_frames = config.get("capture_frames", False)
onnx_fn = path.dirname(
path.abspath(__file__)) + "/data/model360.onnx".replace(
"/", os.path.sep)
class_names_fn = path.dirname(
path.abspath(__file__)) + "/data/face_mask.names".replace(
"/", os.path.sep)
if not path.isfile(onnx_fn):
raise FileNotFoundError(errno.ENOENT, os.strerror(errno.ENOENT),
onnx_fn)
if not path.isfile(class_names_fn):
raise FileNotFoundError(errno.ENOENT, os.strerror(errno.ENOENT),
class_names_fn)
self.class_names = onnx_helper.parse_class_names(class_names_fn)
self.sess_tuple = onnx_helper.get_sess_tuple(onnx_fn)
def generate_anchors(feature_map_sizes_, anchor_sizes_,
anchor_ratios_):
"""
generate anchors.
:param feature_map_sizes_: list of list, for example: [[40,40], [20,20]]
:param anchor_sizes_: list of list, for example: [[0.05, 0.075], [0.1, 0.15]]
:param anchor_ratios_: list of list, for example: [[1, 0.5], [1, 0.5]]
:return:
"""
anchor_bboxes = []
for idx, feature_size in enumerate(feature_map_sizes_):
cx = (np.linspace(0, feature_size[0] - 1, feature_size[0]) +
0.5) / feature_size[0]
cy = (np.linspace(0, feature_size[1] - 1, feature_size[1]) +
0.5) / feature_size[1]
cx_grid, cy_grid = np.meshgrid(cx, cy)
cx_grid_expend = np.expand_dims(cx_grid, axis=-1)
cy_grid_expend = np.expand_dims(cy_grid, axis=-1)
center = np.concatenate((cx_grid_expend, cy_grid_expend),
axis=-1)
num_anchors = len(anchor_sizes_[idx]) + len(
anchor_ratios_[idx]) - 1
center_tiled = np.tile(center, (1, 1, 2 * num_anchors))
anchor_width_heights = []
# different scales with the first aspect ratio
for scale in anchor_sizes_[idx]:
ratio = anchor_ratios_[idx][0] # select the first ratio
width = scale * np.sqrt(ratio)
height = scale / np.sqrt(ratio)
anchor_width_heights.extend([
-width / 2.0, -height / 2.0, width / 2.0, height / 2.0
])
# the first scale, with different aspect ratios (except the first one)
for ratio in anchor_ratios_[idx][1:]:
s1 = anchor_sizes_[idx][0] # select the first scale
width = s1 * np.sqrt(ratio)
height = s1 / np.sqrt(ratio)
anchor_width_heights.extend([
-width / 2.0, -height / 2.0, width / 2.0, height / 2.0
])
bbox_coords = center_tiled + np.array(anchor_width_heights)
bbox_coords_reshape = bbox_coords.reshape((-1, 4))
anchor_bboxes.append(bbox_coords_reshape)
anchor_bboxes = np.concatenate(anchor_bboxes, axis=0)
return anchor_bboxes
feature_map_sizes = [[45, 45], [23, 23], [12, 12], [6, 6], [4, 4]]
# feature_map_sizes = [[33, 33], [17, 17], [9, 9], [5, 5], [3, 3]] #160 için
anchor_sizes = [[0.04, 0.056], [0.08, 0.11], [0.16, 0.22],
[0.32, 0.45], [0.64, 0.72]]
anchor_ratios = [[1, 0.62, 0.42]] * 5
# generate anchors
anchors = generate_anchors(feature_map_sizes, anchor_sizes,
anchor_ratios)
# for inference , the batch size is 1, the model output shape is [1, N, 4],
# so we expand dim for anchors to [1, anchor_num, 4]
self.__anchors_exp = np.expand_dims(anchors, axis=0)