if not os.path.exists(opt.checkpoints_dir):
os.makedirs(opt.checkpoints_dir)
logging.basicConfig(format='%(asctime)s %(message)s', datefmt='%Y/%m/%d %H:%M:%S', \
filename=os.path.join(opt.checkpoints_dir, 'train.log'), level=logging.INFO)
logging.info('======================================================')
logging.info('weight: %f', opt.weight)
print(opt.pretrain_encoder, opt.pretrain_decoder, opt.pretrain_estimater)
logging.info('encoder=%s' % opt.pretrain_encoder)
logging.info('decoder=%s' % opt.pretrain_decoder)
logging.info('estimater=%s' % opt.pretrain_estimater)
if __name__ == '__main__':
#if not os.path.exists(opt.train_record_dir):
# os.system(r"touch{}".format(opt.train_record_dir))
#load data
trainset = ICVL_Loader(opt.dataroot, 'train', opt)
dataset_size = len(trainset)
trainloader = torch.utils.data.DataLoader(trainset,
batch_size=opt.batch_size,
shuffle=True,
num_workers=opt.nThreads)
print('#training point clouds = %d' % len(trainset))
testset = ICVL_Loader(opt.dataroot, 'test', opt)
testloader = torch.utils.data.DataLoader(testset,
batch_size=opt.batch_size,
shuffle=False,
num_workers=opt.nThreads)
print('#testing point clouds = %d' % len(testset))
visualizer = Visualizer(opt)
default=100,
help='weight of estimater, while weight of decoder is 1')
opt = parser.parse_args()
opt.device = torch.device("cuda:%d" %
(opt.gpu_id) if torch.cuda.is_available() else "cpu")
cuda.select_device(opt.gpu_id)
print(opt.pretrain_encoder, opt.pretrain_decoder, opt.pretrain_estimater)
if __name__ == '__main__':
#if not os.path.exists(opt.train_record_dir):
# os.system(r"touch{}".format(opt.train_record_dir))
#load data
testset = ICVL_Loader(opt.dataroot, 'test', opt)
testloader = torch.utils.data.DataLoader(testset,
batch_size=opt.batch_size,
shuffle=False,
num_workers=opt.nThreads)
print('#testing point clouds = %d' % len(testset))
visualizer = Visualizer(opt)
# create model, optionally load pre-trained model
model = Model(opt)
current_en_index = 0
current_es_index = 0
current_de_index = 0
if opt.pretrain_encoder is not None:
model.encoder.load_state_dict(torch.load(opt.pretrain_encoder))
current_en_index = int(