def loadModel(config, modelPath):
model = RFCN_Model(mode="inference",
config=config,
model_dir=os.path.join(ROOT_DIR, "logs"))
if not modelPath:
modelPath = model.find_last()[1]
print("Loading weights from: {}".format(modelPath))
if modelPath and os.path.isfile(modelPath):
# Load trained weights
model.load_weights(modelPath, by_name=True)
else:
raise AssertionError("Model weight file does not exists")
return model
facecolor='none')
ax.add_patch(p)
ax.imshow(image_ori)
plt.savefig(os.path.join(savepath, imgname), bbox_inches='tight')
plt.clf()
if __name__ == '__main__':
ROOT_DIR = os.getcwd()
parser = argparse.ArgumentParser()
parser.add_argument('--loadpath',
required=False,
default="images/",
metavar="evaluate images loadpath",
help="evaluate images loadpath")
parser.add_argument('--savepath',
required=False,
default="result/",
metavar="evaluate images savepath",
help="evaluate images savepath")
config = RFCNNConfig()
args = parser.parse_args()
model = RFCN_Model(mode="inference",
config=config,
model_dir=os.path.join(ROOT_DIR, "logs"))
Test(model, args.loadpath, args.savepath)
if __name__ == '__main__':
ROOT_DIR = os.getcwd()
config = RFCNNConfig()
dataset_train = FashionDataset()
dataset_train.initDB(100000)
dataset_train.prepare()
# Validation dataset
dataset_val = FashionDataset()
dataset_val.initDB(5000, start=100000)
dataset_val.prepare()
model = RFCN_Model(mode="training",
config=config,
model_dir=os.path.join(ROOT_DIR, "logs"))
# This is a hack, bacause the pre-train weights are not fit with dilated ResNet
# model.model.load_weights(
# "resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5", by_name=True, skip_mismatch=True)
try:
model_path = model.find_last()[1]
if model_path is not None:
model.load_weights(model_path, by_name=True)
except Exception as e:
print(e)
print("No checkpoint founded")
# *** This training schedule is an example. Update to your needs ***
def main():
# tensorflow.compat.v1.disable_resource_variables()
# with tensorflow.compat.v1.Session() as ses:
# ses = tensorflow.compat.v1.Session()
# tensorflow.debugging.set_log_device_placement(True)
ROOT_DIR = os.getcwd()
# # inisialisasi config
config = Config()
# inisialisasi dataset training
dataset_train = DeepDriveDataset()
dataset_train.initDB(
"d:/bdd100k/detection_image_train.json",
"D:/Workspace/College/Semester 8/Tugas Akhir/codeR-FCN/bdd100k/images/100k/train",
70000)
dataset_train.prepare()
print(dataset_train.image_ids, "masuk d ong")
# Validation dataset
dataset_val = DeepDriveDataset()
dataset_val.initDB(
"d:/bdd100k/detection_image_val.json",
"D:/Workspace/College/Semester 8/Tugas Akhir/codeR-FCN/bdd100k/images/100k/val",
10000)
dataset_val.prepare()
# siapin model rfcn untuk training
model = RFCN_Model(mode="training",
model_dir=os.path.join(ROOT_DIR, "ModelData"))
# init = tf.global_variables_initializer()
try:
model_path = model.find_last()[1]
print(model_path)
if model_path is not None:
model.load_weights(model_path, by_name=True)
except Exception as e:
print(e)
print("No checkpoint founded")
# K.manual_variable_initialization(True)
model.save("D:/weight.h5")
# tensorflow.compat.v1.enable_resource_variables()
# training begin
# stage 1
model.train(dataset_train,
dataset_val,
learning_rate=config.LEARNING_RATE,
epochs=2,
layers='heads')
model.save("D:/weight.h5")
# Training - Stage 2
# Finetune layers from ResNet stage 4 and up
print("Fine tune Resnet stage 4 and up")
model.train(dataset_train,
dataset_val,
learning_rate=config.LEARNING_RATE,
epochs=40,
layers='4+')
model.save("D:/weight.h5")
# Training - Stage 3
# Fine tune all layers
print("Fine tune all layers")
model.train(dataset_train,
dataset_val,
learning_rate=config.LEARNING_RATE,
epochs=80,
layers='all')
model.save("D:/weight.h5")
# Training - Stage 3
# Fine tune all layers
print("Fine tune all layers")
model.train(dataset_train,
dataset_val,
learning_rate=config.LEARNING_RATE,
epochs=240,
layers='all')
model.save("D:/weight.h5")
ax.add_patch(p)
ax.imshow(image_ori)
plt.savefig(os.path.join(savepath, imgname), bbox_inches='tight')
plt.clf()
if __name__ == '__main__':
ROOT_DIR = os.getcwd()
parser = argparse.ArgumentParser()
parser.add_argument('--loadpath',
required=False,
default="images/",
metavar="evaluate images loadpath",
help="evaluate images loadpath")
parser.add_argument('--savepath',
required=False,
default="result/",
metavar="evaluate images savepath",
help="evaluate images savepath")
config = RFCNNConfig()
args = parser.parse_args()
model = RFCN_Model(mode="inference",
config=config,
model_dir=os.path.join(ROOT_DIR, "logs"),
architecture='resnet101')
Test(model, args.loadpath, args.savepath)
for box in r['rois']:
y1, x1, y2, x2 = box
p = patches.Rectangle((x1, y1),
x2 - x1,
y2 - y1,
linewidth=2,
alpha=0.7,
linestyle="dashed",
edgecolor="red",
facecolor='none')
ax.add_patch(p)
ax.imshow(image_ori)
plt.savefig(os.path.join(savepath, imgname), bbox_inches='tight')
plt.clf()
if __name__ == '__main__':
ROOT_DIR = os.getcwd()
config = RFCNNConfig()
args = parser.parse_args()
model = RFCN_Model(mode="inference",
config=config,
model_dir=os.path.join(
ROOT_DIR,
"/content/drive/My Drive/TugasAkhir/logs"))
Test(model, "/content/100k/test", "/content/result")
plt.clf()
if __name__ == '__main__':
ROOT_DIR = os.getcwd()
parser = argparse.ArgumentParser()
parser.add_argument('--loadpath',
required=False,
default="images/",
metavar="evaluate images loadpath",
help="evaluate images loadpath")
parser.add_argument('--savepath',
required=False,
default="result/",
metavar="evaluate images savepath",
help="evaluate images savepath")
config = RFCNNConfig()
args = parser.parse_args()
modelPath = "/Users/ashish/workspace/ub_coursework/second_semester/Biometrics_and_Image_Analysis/Project/results/"
# model = RFCN_Model(mode="inference", config=config,
# model_dir=os.path.join(ROOT_DIR, "logs") )
model = RFCN_Model(mode="inference", config=config, model_dir=modelPath)
# Test(model, args.loadpath, args.savepath)
loadpath = "images/"
Test(model, loadpath, modelPath)