if __name__ == '__main__':
with open('./config.yml') as f:
cfg = yaml.load(f)
try:
config = cfg[args.arch.upper()]
except AttributeError:
raise ValueError('Unknown architecture: {}'.format(args.arch))
default_boxes = generate_default_boxes(config)
batch_generator, info = create_batch_generator(args.data_dir,
args.data_year,
default_boxes,
config['image_size'],
BATCH_SIZE,
args.num_examples,
mode='test')
try:
ssd = create_ssd(NUM_CLASSES, args.arch, args.pretrained_type,
args.checkpoint_dir, args.checkpoint_path)
except Exception as e:
print(e)
print('The program is exiting...')
sys.exit()
os.makedirs('outputs/images', exist_ok=True)
os.makedirs('outputs/detects', exist_ok=True)
visualizer = ImageVisualizer(info['idx_to_name'],
from anchor import generate_default_boxes
from losses import create_losses
from settings import *
ARCH = 'pre_ssd300-mobilenetv1'
CHECKPOINT_DIR = 'checkpoint/pre_mobilenetv1_lite'
CHECKPOINT_PATH = 'checkpoint/pre_mobilenetv1_lite/ssd_epoch_200.h5'
os.makedirs(CHECKPOINT_DIR, exist_ok=True)
default_boxes = generate_default_boxes(INFO[ARCH])
with tf.device('/device:GPU:0'):
batch_generator, val_generator, info = create_batch_generator(
DATA_DIR, DATA_YEAR, default_boxes,
SIZE, BATCH_SIZE, NUM_BATCHES,
mode='train', augmentation=['flip']) # the patching algorithm is currently causing bottleneck sometimes
dummy = tf.random.normal((1, 300, 300, 3))
ssd = create_pre_ssd_mobilenetv1_lite(weights=None)
pretrained_type = 'specified'
checkpoint_path = CHECKPOINT_PATH
net = init_ssd(ssd, pretrained_type, checkpoint_path)
criterion = create_losses(NEG_RATIO, NUM_CLASSES)
steps_per_epoch = info['length'] // BATCH_SIZE
lr_fn = PiecewiseConstantDecay(
if __name__ == '__main__':
with open('./config.yml') as f:
cfg = yaml.load(f)
try:
config = cfg[args.arch.upper()]
except AttributeError:
raise ValueError('Unknown architecture: {}'.format(args.arch))
default_boxes = generate_default_boxes(config)
batch_generator, info = create_batch_generator(
"./sample_data/images/",
"./sample_data/gt/labels.pkl",
default_boxes,
config['image_size'],
batch_size,
args.num_examples,
mode='test')
try:
# ssd = get_mobilenet_SSD(image_size=(300,300,3), num_classes=num_classes)
ssd = keras.models.load_model("./models/ssd1.h5")
except Exception as e:
print(e)
print('The program is exiting...')
sys.exit()
os.makedirs('outputs/images', exist_ok=True)
os.makedirs('outputs/detects', exist_ok=True)
visualizer = ImageVisualizer(info['idx_to_name'],
with open('./config.yml') as f:
cfg = yaml.load(f)
try:
config = cfg[args.arch.upper()]
except AttributeError:
raise ValueError('Unknown architecture: {}'.format(args.arch))
default_boxes = generate_default_boxes(config)
batch_generator, val_generator, info = create_batch_generator(
args.data_dir,
args.data_year,
default_boxes,
config['image_size'],
args.batch_size,
args.num_batches,
mode='train',
augmentation=args.augment_type,
caching_period=args.caching_period,
snapshot_path=args.snapshot_path
) # the patching algorithm is currently causing bottleneck sometimes
try:
ssd = create_ssd(NUM_CLASSES,
args.arch,
args.pretrained_type,
checkpoint_dir=args.checkpoint_dir)
except Exception as e:
print(e)
print('The program is exiting...')
sys.exit()
with open('./config.yml') as f:
cfg = yaml.load(f)
try:
config = cfg[args.arch.upper()]
except AttributeError:
raise ValueError('Unknown architecture: {}'.format(args.arch))
default_boxes = generate_default_boxes(config)
batch_generator, val_generator, info = create_batch_generator(
"./data/images/",
"./data/gt/labels.pkl",
default_boxes,
config['image_size'],
args.batch_size,
args.num_batches,
mode='train',
augmentation=[
'flip'
]) # the patching algorithm is currently causing bottleneck sometimes
try:
ssd = get_mobilenet_SSD(image_size=(224, 224, 3),
num_classes=NUM_CLASSES)
# ssd = keras.models.load_model("./models/ssd_low_batch.h5")
# this might need to be changed
for layer in ssd.layers:
if 'base' in layer.name:
layer.trainable = False
with open('./config.yml') as f:
cfg = yaml.load(f)
try:
config = cfg[args.arch.upper()]
except AttributeError:
raise ValueError('Unknown architecture: {}'.format(args.arch))
default_boxes = generate_default_boxes(config)
batch_generator, val_generator, info = create_batch_generator(
args.data_dir,
args.data_year,
default_boxes,
config['image_size'],
args.batch_size,
args.num_batches,
mode='train',
augmentation=[
'flip'
]) # the patching algorithm is currently causing bottleneck sometimes
try:
ssd = create_ssd(NUM_CLASSES,
args.arch,
args.pretrained_type,
checkpoint_dir=args.checkpoint_dir)
except Exception as e:
print(e)
print('The program is exiting...')
sys.exit()
if __name__ == '__main__':
# Model의 Checkpoints를 저장할 Directory가 없을 경우 생성한다.
os.makedirs(args.checkpoint_dir, exist_ok=True)
# 실제 SSD 300에 미리 저장되어 있는 Setting값을 가져와서 적용한다.(Anchor, FeatureMapSize 등)
with open('./config.yml') as f:
cfg = yaml.load(f, Loader=yaml.FullLoader)
config = cfg['SSD300']
default_boxes = generate_default_boxes(config)
# voc_data.py에서 설정한 Dataset을 Batch형태로서 가져온다.
batch_generator, val_generator, info = create_batch_generator(
args.data_dir,
default_boxes,
args.batch_size,
args.num_batches,
mode='train')
# 실제 SSD Model을 설정한다. 만약, Training중이던 Model이 있으면 그대로 가져가서 사용할 수 있다.
try:
ssd = create_ssd(NUM_CLASSES,
args.pretrained_type,
checkpoint_dir=args.checkpoint_dir)
except Exception as e:
print(e)
print('The program is exiting...')
sys.exit()
# Hard negative mining을 적용하여 Loss를 구한다.
criterion = create_losses(args.neg_ratio, NUM_CLASSES)
scores = out_scores.numpy()
return boxes, classes, scores
# Model을 정의하게 되고 실제 Detection한 Image의 결과와 Localization, Label 등을 저장하게 된다.
if __name__ == '__main__':
with open('./config.yml') as f:
cfg = yaml.load(f)
config = cfg['SSD300']
default_boxes = generate_default_boxes(config)
batch_generator, info = create_batch_generator(args.data_dir,
default_boxes,
BATCH_SIZE,
args.num_examples,
mode='test')
try:
ssd = create_ssd(NUM_CLASSES, args.pretrained_type,
args.checkpoint_dir, args.checkpoint_path)
except Exception as e:
print(e)
print('The program is exiting...')
sys.exit()
os.makedirs('outputs/images', exist_ok=True)
os.makedirs('outputs/detects', exist_ok=True)
visualizer = ImageVisualizer(info['idx_to_name'],
save_dir='outputs/images')
