help='Start training from an existing model.')
args = parser.parse_args()
# Create the model
model = create_model(existing=args.checkpoint)
# Inform about multi-gpu training
if args.gpus == 1:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpuids
print('Will use GPU ' + args.gpuids)
else:
print('Will use ' + str(args.gpus) + ' GPUs.')
# Data loaders
if args.data == 'nyu':
train_generator, test_generator = get_nyu_train_test_data(args.bs)
# Training session details
runID = str(int(time.time())) + '-n' + str(len(train_generator)) + '-e' + str(args.epochs) + \
'-bs' + str(args.bs) + '-lr' + str(args.lr) + '-' + args.name
outputPath = './models/'
runPath = outputPath + runID
pathlib.Path(runPath).mkdir(parents=True, exist_ok=True)
pathlib.Path(runPath + '/samples').mkdir(parents=True, exist_ok=True)
print('Output: ' + runPath)
# (optional steps)
if True:
# Keep a copy of this training script and calling arguments
with open(__file__, 'r') as training_script:
training_script_content = training_script.read()
default=None,
help='Use non-uniform sampling')
args = parser.parse_args()
# Inform about multi-gpu training
if args.gpus == 1:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpuids
print('Will use GPU ' + args.gpuids)
else:
print('Will use ' + str(args.gpus) + ' GPUs.')
loss = depth_loss_function
# Data loaders
if args.data == 'nyu':
train_generator, test_generator = get_nyu_train_test_data(
args.bs, nus=args.nus_smooth)
if args.data == 'unreal':
train_generator, test_generator = get_unreal_train_test_data(args.bs)
if args.data == 'megadepth':
train_generator, test_generator = get_megadepth_train_test_data(args.bs)
loss = noisy_depth_loss_function
# Create the model
model = create_model(existing=args.checkpoint)
# Training session details
runID = str(int(time.time())) + '-n' + str(len(train_generator)) + '-e' + str(
args.epochs) + '-bs' + str(args.bs) + '-lr' + str(
args.lr) + '-' + args.name
outputPath = './models/'
runPath = outputPath + runID
# Create the model
if args.resnet50: # if want a resnet model
model = create_model_resnet(existing=args.checkpoint)
elif args.efficientnet:
model = create_model_efficientnet(existing=args.checkpoint)
else: # choose densetnet model
if args.dnetVersion == 'small':
model = create_model(existing=args.checkpoint, is121 = True)
if args.dnetVersion == 'medium':
model = create_model( existing=args.checkpoint )
if args.dnetVersion == 'large':
model = create_model( existing=args.checkpoint, is_twohundred=True )
# Data loaders
if args.data == 'nyu': train_generator, test_generator = get_nyu_train_test_data( args.bs )
if args.data == 'redweb': train_generator, test_generator = get_redweb_train_test_data( args.bs )
# Training session details
runID = str(int(time.time())) + '-n' + str(len(train_generator)) + '-e' + str(args.epochs) + '-bs' + str(args.bs) + '-lr' + str(args.lr) + '-' + args.name
outputPath = './models/'
runPath = outputPath + runID
pathlib.Path(runPath).mkdir(parents=True, exist_ok=True)
print('Output: ' + runPath)
# Multi-gpu setup:
basemodel = model
if args.gpus > 1: model = multi_gpu_model(model, gpus=args.gpus)
# Optimizer
optimizer = Adam(lr=args.lr, amsgrad=True)
# Custom object needed for inference and training
custom_objects = {'BilinearUpSampling2D': BilinearUpSampling2D, 'depth_loss_function': None}
print('Loading model...')
# Load model into GPU / CPU
model = load_model(args.model, custom_objects=custom_objects, compile=False)
print('\nModel loaded ({0}).'.format(args.model))
# Input images
# inputs = load_images( glob.glob(args.input) )
# print('\nLoaded ({0}) images of size {1}.'.format(inputs.shape[0], inputs.shape[1:]))
traingen,_ = get_nyu_train_test_data(4)
inputs=traingen[0][0]/255
# Compute results
outputs = predict(model, inputs)
#matplotlib problem on ubuntu terminal fix
#matplotlib.use('TkAgg')
# Display results
viz = display_images(outputs.copy(), inputs.copy())
plt.figure(figsize=(10,5))
plt.imshow(viz)
plt.savefig('test.png')
plt.show()
#------------Test images snippet
import glob
image_list = glob.glob('*.jpg')
test_images = load_images(image_list)
print(test_images.shape)
show_images(test_images)
#------------
'''
Create Model with Decoder
'''
model = create_model()
'''
Create Train and Test Generators
Returns Data_generator objects for Keras
'''
train_generator, test_generator = get_nyu_train_test_data(BATCH_SIZE)
print('\n\nGenerators Ready:', train_generator, test_generator)
optimizer = keras.optimizers.Adam(lr=lr)
model.compile(loss=depth_loss, optimizer=optimizer)
print('\n\nModel Compiled. Ready to train.')
model.fit_generator(train_generator,
validation_data=test_generator,
epochs=EPOCHS,
shuffle=True)
print('Finished Training. Running Inference')