def build_model(cfg, classes='classes'):
yolo = YOLOv4(tiny=cfg.model.tiny)
yolo.classes = classes
yolo.input_size = (cfg.model.input_size, cfg.model.input_size)
yolo.batch_size = cfg.train.batch_size
make_model(yolo)
return yolo
def get_yolo(img_size):
yolo = YOLOv4()
yolo.classes = base_path + "coco.names"
yolo.input_size = img_size
yolo.make_model()
yolo.load_weights(base_path + "yolov4.weights", weights_type="yolo")
return yolo
def load_model():
from yolov4.tf import YOLOv4
""" Load the model from disk. This should be called before the predict API """
start_time = time.time()
# LOGGER.info("Start loading the YoloV4 model")
YoloV4Wrapper.yolo = YOLOv4()
YoloV4Wrapper.yolo.classes = YOLOV4_CLASSES
YoloV4Wrapper.yolo.make_model()
YoloV4Wrapper.yolo.load_weights(YOLOV4_WEIGHTS, weights_type="yolo")
YoloV4Wrapper.is_model_loaded = True
exec_time = time.time() - start_time
def __init__(self):
self.yolo = YOLOv4()
self.yolo.classes = './coco.names'
self.yolo.make_model()
self.yolo.load_weights("./model/yolov4-custom_last.weights",
weights_type="yolo")
self.config = Cfg.load_config()
self.config['weights'] = './model/transformerocr.pth'
self.config['predictor']['beamsearch'] = False
self.config['device'] = 'cpu'
self.detector = Predictor(self.config)
self.classes = ['id', 'name', 'dmy', 'add1', 'add2']
self.res = dict.fromkeys(self.classes, '')
import cv2
from yolov4.tf import YOLOv4
yolo = YOLOv4()
yolo.config.parse_names("coco.names")
yolo.config.parse_cfg("yolov4-tiny.cfg") #Remember to change weight source both for .cfg and .weights
yolo.make_model()
yolo.load_weights("yolov4-tiny.weights", weights_type="yolo")
yolo.summary(summary_type="yolo")
yolo.summary()
##yolo.inference(media_path="download.jpeg")
##yolo.inference(media_path="road.mp4", is_image=False)
#To find camera sources available, run ls -ltrh /dev/video*
yolo.inference(
"/dev/video6", #video0 is webcam, video6 is realsense RGB
is_image=False,
cv_apiPreference=cv2.CAP_V4L2,
cv_frame_size=(640, 480),
cv_fourcc="YUYV",
)
from tensorflow.keras import callbacks, optimizers
from yolov4.tf import SaveWeightsCallback, YOLOv4
import time
yolo = YOLOv4(tiny=True)
yolo.classes = "./yolofile/coco.names"
yolo.input_size = 608
yolo.batch_size = 32
yolo.make_model()
yolo.load_weights("./yolofile/yolov4-tiny.conv.29", weights_type="yolo")
train_data_set = yolo.load_dataset(
"./yolofile/train2017.txt",
image_path_prefix="/Users/yu.duan/Documents/data/coco_dataset/train2017",
label_smoothing=0.05)
val_data_set = yolo.load_dataset(
"./yolofile/val2017.txt",
image_path_prefix=" /Users/yu.duan/Documents/data/coco_dataset/val2017",
training=False)
epochs = 400
lr = 1e-4
optimizer = optimizers.Adam(learning_rate=lr)
yolo.compile(optimizer=optimizer, loss_iou_type="ciou")
def lr_scheduler(epoch):
if epoch < int(epochs * 0.5):
return lr
"""
export model as keras, onnx and edge tpu model
onnx export requires tensorflow-onnx package
"""
import tensorflow as tf
import subprocess
from yolov4.tf import YOLOv4
MODEL_PATH = "tiny_yolov4_relu/"
WEIGHT_PATH = "yolov4-tiny-relu.weights"
DATASET_PATH = "path/to/dataset"
# create model
yolov4 = YOLOv4(tiny=True, tpu=True)
yolov4.classes = "dataset/coco.names"
yolov4.make_model(activation1="relu")
yolov4.load_weights(WEIGHT_PATH, weights_type="yolo")
# save as keras model
yolov4.model.save(MODEL_PATH)
# save as onnx model
try:
subprocess.run(
[
"python",
"-m",
"tf2onnx.convert",
"--opset",
parser.add_argument('--save-video',
action='store_true',
help='Create a video file with the analysis result.')
return parser.parse_args()
if __name__ == '__main__':
args = parse_args()
# Init the object tracker
tracker = CentroidTracker()
# Init the model
yolo = YOLOv4(isTiny(bytes(args.weight, encoding="utf-8")))
yolo.classes = "../darknet_files/dataset/classes.names"
yolo.make_model()
yolo.load_weights(bytes(args.weight, encoding="utf-8"),
weights_type="yolo")
cap = cv.VideoCapture(args.video)
bar = Bar('Processing Frames', max=int(cap.get(cv.CAP_PROP_FRAME_COUNT)))
width = int(cap.get(3))
height = int(cap.get(4))
if (not cap.isOpened()):
print('Error opening video stream!')
exit()
print(f'\nAnalysing {args.video}...')
