def test_yolov3_predict():
y = YOLO_V3(
utils.load_json('configs/config_classification_flowers17.json'))
while True:
y.predict_classification(
image_path="/Users/shidanlifuhetian/All/data/flowers17_tsycnh/test"
)
from yolo_model import YOLO_V3
from utils import load_json
from data_generator import St_Generator
if __name__ == "__main__":
# this is for detection training
config = load_json('configs/config_detection_defects_winK40.json')
gen = St_Generator(config, phase="train", shuffle=True)
val_gen = St_Generator(config, phase="test")
yolo = YOLO_V3(config=config)
if yolo.config['train']['mode'] == "transfer learning":
print("transfer learning")
for layer in yolo.backbone.layers:
layer.trainable = False
if yolo.config['train']['mode'] == "fine tune":
print("fine tune")
trainable_point = False
for layer in yolo.backbone.layers:
if layer.name == "conv2d_44": # 3x3/2 filters:1024
trainable_point = True
if trainable_point == True:
layer.trainable = True
else:
layer.trainable = False
yolo.train_detection(gen, val_gen)
def test_yolov3_classification():
y = YOLO_V3(utils.load_json('./unit_test/test_config2.json'))