def main():
# Get args from command line
args = get_args_from_command_line()
if args.gpu_id is not None:
cfg.CONST.DEVICE = args.gpu_id
if not args.randomize:
np.random.seed(cfg.CONST.RNG_SEED)
if args.batch_size is not None:
cfg.CONST.BATCH_SIZE = args.batch_size
if args.iter is not None:
cfg.TRAIN.NUM_ITERATION = args.iter
if args.out_path is not None:
cfg.DIR.OUT_PATH = args.out_path
if args.weights is not None:
cfg.CONST.WEIGHTS = args.weights
cfg.TRAIN.RESUME_TRAIN = True
cfg.TRAIN.INITIAL_ITERATION = int(args.init_iter)
# Print config
print('Use config:')
pprint(cfg)
# Set GPU to use
theano.gpuarray.use(cfg.CONST.DEVICE)
# Start train/test process
if not args.test:
train_net(cfg)
else:
test_net(cfg)
def main():
# Get args from command line
args = get_args_from_command_line()
if args.gpu_id is not None:
cfg.CONST.DEVICE = args.gpu_id
if args.weights is not None:
cfg.CONST.WEIGHTS = args.weights
# Print config
print('Use config:')
pprint(cfg)
# Set GPU to use
os.environ["CUDA_VISIBLE_DEVICES"] = cfg.CONST.DEVICE
# Start train/test process
if not args.test and not args.inference:
train_net(cfg)
else:
if 'WEIGHTS' not in cfg.CONST or not os.path.exists(cfg.CONST.WEIGHTS):
logging.error('Please specify the file path of checkpoint.')
sys.exit(2)
if args.test:
test_net(cfg)
else:
inference_net(cfg)
def main():
# Get args from command line
args = get_args_from_command_line()
if args.gpu_id is not None:
cfg.CONST.DEVICE = args.gpu_id
if args.weights is not None:
cfg.CONST.WEIGHTS = args.weights
# Print config
print('Use config:')
pprint(cfg)
# f_runner.write(str(cfg))
# Set GPU to use
os.environ["CUDA_VISIBLE_DEVICES"] = cfg.CONST.DEVICE
# Start train/test process
if not args.test and not args.inference:
train_net(cfg)
else:
'''
if 'WEIGHTS' not in cfg.CONST or not os.path.exists(cfg.CONST.WEIGHTS):
logging.error('Please specify the file path of checkpoint.')
sys.exit(2)
'''
if 'WEIGHTS' not in cfg.CONST:
logging.error('Please specify the file path of checkpoint.1')
sys.exit(2)
if not os.path.exists(cfg.CONST.WEIGHTS):
logging.error('Please specify the file path of checkpoint.2')
sys.exit(2)
if args.test:
# test_net(cfg)
path = '/raid/wuruihai/GRNet_FILES/tb_log'
test_writer = SummaryWriter(path)
test_net(cfg, test_writer=test_writer)
if args.test_KITTI:
path = '/raid/wuruihai/GRNet_FILES/tb_log'
test_writer = SummaryWriter(path)
test_net_KITTI(cfg, test_writer=test_writer)
else:
inference_net(cfg)
def main():
# Get args from command line
args = get_args_from_command_line()
# Read the experimental config
exec(compile(open(args.cfg_file, "rb").read(), args.cfg_file, 'exec'))
cfg = locals()['__C']
pprint(cfg)
# Parse runtime arguments
if args.gpu_id is not None:
os.environ["CUDA_VISIBLE_DEVICES"] = args.gpu_id
if not args.randomize:
random.seed(cfg.CONST.RNG_SEED)
np.random.seed(cfg.CONST.RNG_SEED)
torch.manual_seed(cfg.CONST.RNG_SEED)
torch.cuda.manual_seed(cfg.CONST.RNG_SEED)
torch.cuda.manual_seed_all(cfg.CONST.RNG_SEED)
# References: https://pytorch.org/docs/stable/notes/randomness.html
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
if args.exp_name is not None:
cfg.CONST.EXP_NAME = args.exp_name
if args.weights is not None:
cfg.CONST.WEIGHTS = args.weights
# Start train/test process
if not args.test and not args.inference:
# Make sure cfg.TRAIN.NETWORK in ['RMNet', 'TinyFlowNet']
if cfg.TRAIN.NETWORK not in ['RMNet', 'TinyFlowNet']:
logging.error(
'Please make sure cfg.TRAIN.NETWORK in ["RMNet", "TinyFlowNet"].'
)
sys.exit(1)
train_net(cfg)
else:
if 'WEIGHTS' not in cfg.CONST or not os.path.exists(cfg.CONST.WEIGHTS):
logging.error('Please specify the file path of checkpoint.')
sys.exit(2)
if args.test:
test_net(cfg)
else:
inference_net(cfg)
def main():
# Get args from command line
args = get_args_from_command_line()
if args.gpu_id is not None:
cfg.CONST.DEVICE = args.gpu_id
if not args.randomize:
np.random.seed(cfg.CONST.RNG_SEED)
if args.batch_size is not None:
cfg.CONST.BATCH_SIZE = args.batch_size
if args.epoch is not None:
cfg.TRAIN.NUM_EPOCHES = args.epoch
if args.dataset is not None:
cfg.DATASET.DATASET_NAME = args.dataset
if cfg.DATASET.DATASET_NAME not in cfg.DATASETS:
cfg.DATASET.CENTER_BIAS = cfg.DATASETS.TEST.CENTER_BIAS
else:
cfg.DATASET.CENTER_BIAS = cfg.DATASETS.SALICON.CENTER_BIAS
if args.out_path is not None:
cfg.DIR.OUT_PATH = args.out_path
if args.weights is not None:
cfg.CONST.WEIGHTS = args.weights
if not args.test:
cfg.TRAIN.RESUME_TRAIN = True
# print config
print('Use config:')
pprint(cfg)
# Set GPU to use
if type(cfg.CONST.DEVICE) == str:
os.environ["CUDA_VISIBLE_DEVICES"] = cfg.CONST.DEVICE
# Start train/test process
if not args.test:
train_net(cfg)
else:
if 'WEIGHTS' in cfg.CONST and os.path.exists(cfg.CONST.WEIGHTS):
test_net(cfg)
else:
print('[FATAL] %s Please specify the file path of checkpoint.' %
(dt.now()))
sys.exit(2)
import dragon.memonger as opt
opt.ShareGrads()
cfg.DATA_DIR = 'data/quad'
if __name__ == '__main__':
# fix the random seeds (numpy and caffe) for reproducibility
np.random.seed(cfg.RNG_SEED)
caffe.set_random_seed(cfg.RNG_SEED)
# setup caffe
caffe.set_mode_gpu()
caffe.set_device(gpu_id)
# setup database
cfg.DATABASE = imdb_name
imdb = get_imdb(imdb_name)
print 'Database({}): {} images will be used to train.'.format(
cfg.DATABASE, imdb.db_size)
output_dir = osp.abspath(osp.join(cfg.ROOT_DIR, 'output', imdb_name))
print 'Output will be saved to `{:s}`'.format(output_dir)
# train net
train_net(solver_txt,
output_dir,
pretrained_model=pretrained_model,
snapshot_model=snapshot_model,
start_iter=start_iter,
max_iters=max_iters,
warm_up=warm_up)
cfg.GPU_ID = args.gpu_id
print('Using config:')
pprint.pprint(cfg)
if not args.randomize:
# fix the random seeds (numpy and caffe) for reproducibility
np.random.seed(cfg.RNG_SEED)
caffe.set_random_seed(cfg.RNG_SEED)
# set up caffe
caffe.set_mode_gpu()
caffe.set_device(args.gpu_id)
imdb, roidb = get_roidb(args.imdb_name)
print '{:d} roidb entries'.format(len(roidb))
print("num_classes", imdb.num_classes)
print(imdb.roidb_num_bboxes_at(-1))
print(roidb[0])
sys.exit()
output_dir = get_output_dir(imdb)
print 'Output will be saved to `{:s}`'.format(output_dir)
train_net(args.solver,
roidb,
output_dir,
pretrained_model=args.pretrained_model,
max_iters=args.max_iters)
print(sample.keys())
print(sample['flipped'])
"""
al_net = None
if args.al_net is not None and args.al_def is not None:
al_net = caffe.Net(args.al_def, caffe.TEST, weights=args.al_net)
al_net.name = "al_" + os.path.splitext(os.path.basename(
args.al_def))[0]
print '{:d} roidb entries'.format(len(roidb))
print 'Output will be saved to `{:s}`'.format(output_dir)
train_net(args.solver,
roidb,
output_dir,
solver_state=args.solver_state,
pretrained_model=args.pretrained_model,
max_iters=args.max_iters,
al_net=al_net)
'''
argparse.ArgumentParser
Input: (description='Train a Fast R-CNN network'), Output: parser
get_repo_imdb
input: (imdb_name), output: imdb
get_training_roidb
input: (imdb), output: roidb
get_roidb
input: (args.imdb_name), output: imdb, roidb
if __name__ == '__main__':
args = parse_args()
print('Called with args:')
print(args)
if args.cfg_file is not None:
cfg_from_file(args.cfg_file)
if args.exp_dir is not None:
cfg.EXP_DIR = args.exp_dir
cfg.GPU_ID = args.gpu_id
print('Using config:')
pprint.pprint(cfg)
np.random.seed(cfg.RNG_SEED)
caffe.set_random_seed(cfg.RNG_SEED)
random.seed(cfg.RNG_SEED)
# set up caffe
caffe.set_mode_gpu()
caffe.set_device(args.gpu_id)
time_suffix = 'train_' + datetime.datetime.now().strftime("%d_%m_%Y_%H_%M")
output_dir = get_output_dir(time_suffix, None)
print('Output will be saved to `{:s}`'.format(output_dir))
train_net(output_dir)