def main():
args = parse_args()
if args.out is not None and not args.out.endswith(('.pkl', '.pickle')):
raise ValueError('The output file must be a pkl file.')
cfg = mmcv.Config.fromfile(args.config)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.model.pretrained = None
cfg.data.test.test_mode = True
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
imgs_per_gpu = cfg.data.test.pop('imgs_per_gpu', 1)
# build the dataloader
# TODO: support multiple images per gpu (only minor changes are needed)
dataset = get_dataset(cfg.data.test)
data_loader = build_dataloader(dataset,
imgs_per_gpu=imgs_per_gpu,
workers_per_gpu=cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False)
# build the model and load checkpoint
model = build_detector(cfg.model, train_cfg=None, test_cfg=cfg.test_cfg)
checkpoint = load_checkpoint(model, args.checkpoint, map_location='cpu')
# old versions did not save class info in checkpoints, this walkaround is
# for backward compatibility
if 'CLASSES' in checkpoint['meta']:
model.CLASSES = checkpoint['meta']['CLASSES']
else:
model.CLASSES = dataset.CLASSES
if not distributed:
model = MMDataParallel(model, device_ids=[0])
outputs = single_gpu_test(model, data_loader, args.show, args.log_dir)
else:
model = MMDistributedDataParallel(model.cuda())
outputs = multi_gpu_test(model, data_loader, args.tmpdir)
rank, _ = get_dist_info()
if args.out and rank == 0:
print('\nwriting results to {}'.format(args.out))
mmcv.dump(outputs, args.out)
eval_types = args.eval
if args.csv:
csv_path = (args.out).replace('.pkl', '.csv')
print('\nwriting results as csv to {}'.format(csv_path))
convert_output_to_csv(dataset, outputs, csv_path)
if eval_types:
print('Starting evaluate {}'.format(' and '.join(eval_types)))
if eval_types == ['proposal_fast']:
result_file = args.out
coco_eval(result_file, eval_types, dataset.coco)
else:
if not isinstance(outputs[0], dict):
result_file = args.out + '.json'
results2json(dataset, outputs, result_file)
coco_eval(result_file, eval_types, dataset.coco)
else:
for name in outputs[0]:
print('\nEvaluating {}'.format(name))
outputs_ = [out[name] for out in outputs]
result_file = args.out + '.{}.json'.format(name)
results2json(dataset, outputs_, result_file)
coco_eval(result_file, eval_types, dataset.coco)
def main():
args = parse_args()
cfg = mmcv.Config.fromfile(args.config)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.model.pretrained = None
cfg.data.test.test_mode = True
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# build the dataloader
dataset = build_dataset(cfg.data.test)
data_loader = build_dataloader(dataset,
samples_per_gpu=cfg.data.samples_per_gpu,
workers_per_gpu=cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False,
round_up=False)
# build the model and load checkpoint
model = build_classifier(cfg.model)
fp16_cfg = cfg.get('fp16', None)
if fp16_cfg is not None:
wrap_fp16_model(model)
_ = load_checkpoint(model, args.checkpoint, map_location='cpu')
if not distributed:
model = MMDataParallel(model, device_ids=[0])
outputs = single_gpu_test(model, data_loader)
else:
model = MMDistributedDataParallel(
model.cuda(),
device_ids=[torch.cuda.current_device()],
broadcast_buffers=False)
outputs = multi_gpu_test(model, data_loader, args.tmpdir,
args.gpu_collect)
rank, _ = get_dist_info()
if rank == 0:
nums = []
results = {}
for output in outputs:
nums.append(output['num_samples'].item())
for topk, v in output['accuracy'].items():
if topk not in results:
results[topk] = []
results[topk].append(v.item())
assert sum(nums) == len(dataset)
for topk, accs in results.items():
avg_acc = np.average(accs, weights=nums)
print(f'\n{topk} accuracy: {avg_acc:.2f}')
if args.out and rank == 0:
print(f'\nwriting results to {args.out}')
mmcv.dump(outputs, args.out)
def main():
args = parse_args()
assert args.out or args.eval or args.format_only or args.show \
or args.show_dir, \
('Please specify at least one operation (save/eval/format/show the '
'results / save the results) with the argument "--out", "--eval"'
', "--format-only", "--show" or "--show-dir"')
if args.eval and args.format_only:
raise ValueError('--eval and --format_only cannot be both specified')
if args.out is not None and not args.out.endswith(('.pkl', '.pickle')):
raise ValueError('The output file must be a pkl file.')
cfg = Config.fromfile(args.config)
if args.cfg_options is not None:
cfg.merge_from_dict(args.cfg_options)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.model.pretrained = None
if cfg.model.get('neck'):
if isinstance(cfg.model.neck, list):
for neck_cfg in cfg.model.neck:
if neck_cfg.get('rfp_backbone'):
if neck_cfg.rfp_backbone.get('pretrained'):
neck_cfg.rfp_backbone.pretrained = None
elif cfg.model.neck.get('rfp_backbone'):
if cfg.model.neck.rfp_backbone.get('pretrained'):
cfg.model.neck.rfp_backbone.pretrained = None
# in case the test dataset is concatenated
samples_per_gpu = 1
if isinstance(cfg.data.test, dict):
cfg.data.test.test_mode = True
samples_per_gpu = cfg.data.test.pop('samples_per_gpu', 1)
if samples_per_gpu > 1:
# Replace 'ImageToTensor' to 'DefaultFormatBundle'
cfg.data.test.pipeline = replace_ImageToTensor(
cfg.data.test.pipeline)
elif isinstance(cfg.data.test, list):
for ds_cfg in cfg.data.test:
ds_cfg.test_mode = True
samples_per_gpu = max(
[ds_cfg.pop('samples_per_gpu', 1) for ds_cfg in cfg.data.test])
if samples_per_gpu > 1:
for ds_cfg in cfg.data.test:
ds_cfg.pipeline = replace_ImageToTensor(ds_cfg.pipeline)
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
rank, _ = get_dist_info()
# allows not to create
if args.work_dir is not None and rank == 0:
mmcv.mkdir_or_exist(osp.abspath(args.work_dir))
timestamp = time.strftime('%Y%m%d_%H%M%S', time.localtime())
json_file = osp.join(args.work_dir, f'eval_{timestamp}.json')
# build the dataloader
dataset = build_dataset(cfg.data.test)
data_loader = build_dataloader(dataset,
samples_per_gpu=samples_per_gpu,
workers_per_gpu=cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False)
# build the model and load checkpoint
cfg.model.train_cfg = None
model = build_detector(cfg.model, test_cfg=cfg.get('test_cfg'))
fp16_cfg = cfg.get('fp16', None)
if fp16_cfg is not None:
wrap_fp16_model(model)
checkpoint = load_checkpoint(model, args.checkpoint, map_location='cpu')
if args.fuse_conv_bn:
model = fuse_conv_bn(model)
# old versions did not save class info in checkpoints, this walkaround is
# for backward compatibility
if 'CLASSES' in checkpoint.get('meta', {}):
model.CLASSES = checkpoint['meta']['CLASSES']
else:
model.CLASSES = dataset.CLASSES
if not distributed:
model = MMDataParallel(model, device_ids=[0])
outputs = single_gpu_test(model, data_loader, args.show, args.show_dir,
args.show_score_thr)
else:
model = MMDistributedDataParallel(
model.cuda(),
device_ids=[torch.cuda.current_device()],
broadcast_buffers=False)
outputs = multi_gpu_test(model, data_loader, args.tmpdir,
args.gpu_collect)
rank, _ = get_dist_info()
if rank == 0:
if args.out:
print(f'\nwriting results to {args.out}')
mmcv.dump(outputs, args.out)
kwargs = {} if args.eval_options is None else args.eval_options
if args.format_only:
dataset.format_results(outputs, **kwargs)
if args.eval:
eval_kwargs = cfg.get('evaluation', {}).copy()
# hard-code way to remove EvalHook args
for key in [
'interval', 'tmpdir', 'start', 'gpu_collect', 'save_best',
'rule'
]:
eval_kwargs.pop(key, None)
eval_kwargs.update(dict(metric=args.eval, **kwargs))
metric = dataset.evaluate(outputs, **eval_kwargs)
print(metric)
metric_dict = dict(config=args.config, metric=metric)
if args.work_dir is not None and rank == 0:
mmcv.dump(metric_dict, json_file)
def main():
args = parse_args()
assert args.out or args.show, \
('Please specify at least one operation (save or show the results) '
'with the argument "--out" or "--show"')
if args.out is not None and not args.out.endswith(('.pkl', '.pickle')):
raise ValueError('The output file must be a pkl file.')
cfg = mmcv.Config.fromfile(args.config)
# import modules from string list.
if cfg.get('custom_imports', None):
from mmcv.utils import import_modules_from_strings
import_modules_from_strings(**cfg['custom_imports'])
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.model.pretrained = None
cfg.data.test.test_mode = True
if args.workers == 0:
args.workers = cfg.data.workers_per_gpu
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# set random seeds
if args.seed is not None:
set_random_seed(args.seed)
if 'all' in args.corruptions:
corruptions = [
'gaussian_noise', 'shot_noise', 'impulse_noise', 'defocus_blur',
'glass_blur', 'motion_blur', 'zoom_blur', 'snow', 'frost', 'fog',
'brightness', 'contrast', 'elastic_transform', 'pixelate',
'jpeg_compression', 'speckle_noise', 'gaussian_blur', 'spatter',
'saturate'
]
elif 'benchmark' in args.corruptions:
corruptions = [
'gaussian_noise', 'shot_noise', 'impulse_noise', 'defocus_blur',
'glass_blur', 'motion_blur', 'zoom_blur', 'snow', 'frost', 'fog',
'brightness', 'contrast', 'elastic_transform', 'pixelate',
'jpeg_compression'
]
elif 'noise' in args.corruptions:
corruptions = ['gaussian_noise', 'shot_noise', 'impulse_noise']
elif 'blur' in args.corruptions:
corruptions = [
'defocus_blur', 'glass_blur', 'motion_blur', 'zoom_blur'
]
elif 'weather' in args.corruptions:
corruptions = ['snow', 'frost', 'fog', 'brightness']
elif 'digital' in args.corruptions:
corruptions = [
'contrast', 'elastic_transform', 'pixelate', 'jpeg_compression'
]
elif 'holdout' in args.corruptions:
corruptions = ['speckle_noise', 'gaussian_blur', 'spatter', 'saturate']
elif 'None' in args.corruptions:
corruptions = ['None']
args.severities = [0]
else:
corruptions = args.corruptions
rank, _ = get_dist_info()
aggregated_results = {}
for corr_i, corruption in enumerate(corruptions):
aggregated_results[corruption] = {}
for sev_i, corruption_severity in enumerate(args.severities):
# evaluate severity 0 (= no corruption) only once
if corr_i > 0 and corruption_severity == 0:
aggregated_results[corruption][0] = \
aggregated_results[corruptions[0]][0]
continue
test_data_cfg = copy.deepcopy(cfg.data.test)
# assign corruption and severity
if corruption_severity > 0:
corruption_trans = dict(type='Corrupt',
corruption=corruption,
severity=corruption_severity)
# TODO: hard coded "1", we assume that the first step is
# loading images, which needs to be fixed in the future
test_data_cfg['pipeline'].insert(1, corruption_trans)
# print info
print(f'\nTesting {corruption} at severity {corruption_severity}')
# build the dataloader
# TODO: support multiple images per gpu
# (only minor changes are needed)
dataset = build_dataset(test_data_cfg)
data_loader = build_dataloader(dataset,
samples_per_gpu=1,
workers_per_gpu=args.workers,
dist=distributed,
shuffle=False)
# build the model and load checkpoint
model = build_detector(cfg.model,
train_cfg=None,
test_cfg=cfg.test_cfg)
fp16_cfg = cfg.get('fp16', None)
if fp16_cfg is not None:
wrap_fp16_model(model)
checkpoint = load_checkpoint(model,
args.checkpoint,
map_location='cpu')
# old versions did not save class info in checkpoints,
# this walkaround is for backward compatibility
if 'CLASSES' in checkpoint['meta']:
model.CLASSES = checkpoint['meta']['CLASSES']
else:
model.CLASSES = dataset.CLASSES
if not distributed:
model = MMDataParallel(model, device_ids=[0])
show_dir = args.show_dir
if show_dir is not None:
show_dir = osp.join(show_dir, corruption)
show_dir = osp.join(show_dir, corruption_severity)
if not osp.exists(show_dir):
osp.makedirs(show_dir)
outputs = single_gpu_test(model, data_loader, args.show,
show_dir, args.show_score_thr)
else:
model = MMDistributedDataParallel(
model.cuda(),
device_ids=[torch.cuda.current_device()],
broadcast_buffers=False)
outputs = multi_gpu_test(model, data_loader, args.tmpdir)
if args.out and rank == 0:
eval_results_filename = (osp.splitext(args.out)[0] +
'_results' +
osp.splitext(args.out)[1])
mmcv.dump(outputs, args.out)
eval_types = args.eval
if cfg.dataset_type == 'VOCDataset':
if eval_types:
for eval_type in eval_types:
if eval_type == 'bbox':
test_dataset = mmcv.runner.obj_from_dict(
cfg.data.test, datasets)
logger = 'print' if args.summaries else None
mean_ap, eval_results = \
voc_eval_with_return(
args.out, test_dataset,
args.iou_thr, logger)
aggregated_results[corruption][
corruption_severity] = eval_results
else:
print('\nOnly "bbox" evaluation \
is supported for pascal voc')
else:
if eval_types:
print(f'Starting evaluate {" and ".join(eval_types)}')
if eval_types == ['proposal_fast']:
result_file = args.out
else:
if not isinstance(outputs[0], dict):
result_files = dataset.results2json(
outputs, args.out)
else:
for name in outputs[0]:
print(f'\nEvaluating {name}')
outputs_ = [out[name] for out in outputs]
result_file = args.out
+ f'.{name}'
result_files = dataset.results2json(
outputs_, result_file)
eval_results = coco_eval_with_return(
result_files, eval_types, dataset.coco)
aggregated_results[corruption][
corruption_severity] = eval_results
else:
print('\nNo task was selected for evaluation;'
'\nUse --eval to select a task')
# save results after each evaluation
mmcv.dump(aggregated_results, eval_results_filename)
if rank == 0:
# print filan results
print('\nAggregated results:')
prints = args.final_prints
aggregate = args.final_prints_aggregate
if cfg.dataset_type == 'VOCDataset':
get_results(eval_results_filename,
dataset='voc',
prints=prints,
aggregate=aggregate)
else:
get_results(eval_results_filename,
dataset='coco',
prints=prints,
aggregate=aggregate)
def main():
args = parse_args()
cfg = Config.fromfile(args.config)
if args.cfg_options is not None:
cfg.merge_from_dict(args.cfg_options)
# import modules from string list.
if cfg.get('custom_imports', None):
from mmcv.utils import import_modules_from_strings
import_modules_from_strings(**cfg['custom_imports'])
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
# work_dir is determined in this priority: CLI > segment in file > filename
if args.work_dir is not None:
# update configs according to CLI args if args.work_dir is not None
cfg.work_dir = args.work_dir
elif cfg.get('work_dir', None) is None:
# use config filename as default work_dir if cfg.work_dir is None
cfg.work_dir = osp.join('./work_dirs',
osp.splitext(osp.basename(args.config))[0])
if args.resume_from is not None:
cfg.resume_from = args.resume_from
if args.gpu_ids is not None:
cfg.gpu_ids = args.gpu_ids
else:
cfg.gpu_ids = range(1) if args.gpus is None else range(args.gpus)
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# re-set gpu_ids with distributed training mode
_, world_size = get_dist_info()
cfg.gpu_ids = range(world_size)
# create work_dir
mmcv.mkdir_or_exist(osp.abspath(cfg.work_dir))
# dump config
cfg.dump(osp.join(cfg.work_dir, osp.basename(args.config)))
# init the logger before other steps
timestamp = time.strftime('%Y%m%d_%H%M%S', time.localtime())
log_file = osp.join(cfg.work_dir, f'{timestamp}.log')
logger = get_root_logger(log_file=log_file, log_level=cfg.log_level)
# init the meta dict to record some important information such as
# environment info and seed, which will be logged
meta = dict()
# log env info
env_info_dict = collect_env()
env_info = '\n'.join([(f'{k}: {v}') for k, v in env_info_dict.items()])
dash_line = '-' * 60 + '\n'
logger.info('Environment info:\n' + dash_line + env_info + '\n' +
dash_line)
meta['env_info'] = env_info
meta['config'] = cfg.pretty_text
# log some basic info
logger.info(f'Distributed training: {distributed}')
logger.info(f'Config:\n{cfg.pretty_text}')
# set random seeds
if args.seed is not None:
logger.info(f'Set random seed to {args.seed}, '
f'deterministic: {args.deterministic}')
set_random_seed(args.seed, deterministic=args.deterministic)
cfg.seed = args.seed
meta['seed'] = args.seed
meta['exp_name'] = osp.basename(args.config)
model = build_detector(
cfg.model,
train_cfg=cfg.get('train_cfg'),
test_cfg=cfg.get('test_cfg'))
datasets = [build_dataset(cfg.data.train)]
if len(cfg.workflow) == 2:
val_dataset = copy.deepcopy(cfg.data.val)
val_dataset.pipeline = cfg.data.train.pipeline
datasets.append(build_dataset(val_dataset))
if cfg.checkpoint_config is not None:
# save mmdet version, config file content and class names in
# checkpoints as meta data
cfg.checkpoint_config.meta = dict(
mmdet_version=__version__ + get_git_hash()[:7],
CLASSES=datasets[0].CLASSES)
# add an attribute for visualization convenience
model.CLASSES = datasets[0].CLASSES
train_detector(
model,
datasets,
cfg,
distributed=distributed,
validate=(not args.no_validate),
timestamp=timestamp,
meta=meta)
def main():
args = parse_args()
cfg = Config.fromfile(args.config)
cfg.merge_from_dict(args.cfg_options)
# Load output_config from cfg
output_config = cfg.get('output_config', {})
if args.out:
# Overwrite output_config from args.out
output_config = Config._merge_a_into_b(dict(out=args.out),
output_config)
# Load eval_config from cfg
eval_config = cfg.get('eval_config', {})
if args.eval:
# Overwrite eval_config from args.eval
eval_config = Config._merge_a_into_b(dict(metrics=args.eval),
eval_config)
if args.eval_options:
# Add options from args.eval_options
eval_config = Config._merge_a_into_b(args.eval_options, eval_config)
assert output_config or eval_config, \
('Please specify at least one operation (save or eval the '
'results) with the argument "--out" or "--eval"')
dataset_type = cfg.data.test.type
if output_config.get('out', None):
out = output_config['out']
# make sure the dirname of the output path exists
mmcv.mkdir_or_exist(osp.dirname(out))
_, suffix = osp.splitext(out)
if dataset_type == 'AVADataset':
assert suffix[1:] == 'csv', ('For AVADataset, the format of the '
'output file should be csv')
else:
assert suffix[1:] in file_handlers, (
'The format of the output '
'file should be json, pickle or yaml')
# set cudnn benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.data.test.test_mode = True
if cfg.test_cfg is None:
cfg.test_cfg = dict(average_clips=args.average_clips)
else:
# You can set average_clips during testing, it will override the
# original settting
if args.average_clips is not None:
cfg.test_cfg.average_clips = args.average_clips
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# The flag is used to register module's hooks
cfg.setdefault('module_hooks', [])
# build the dataloader
dataset = build_dataset(cfg.data.test, dict(test_mode=True))
dataloader_setting = dict(videos_per_gpu=cfg.data.get('videos_per_gpu', 1),
workers_per_gpu=cfg.data.get(
'workers_per_gpu', 1),
dist=distributed,
shuffle=False)
dataloader_setting = dict(dataloader_setting,
**cfg.data.get('test_dataloader', {}))
data_loader = build_dataloader(dataset, **dataloader_setting)
# build the model and load checkpoint
model = build_model(cfg.model, train_cfg=None, test_cfg=cfg.test_cfg)
register_module_hooks(model.backbone, cfg.module_hooks)
fp16_cfg = cfg.get('fp16', None)
if fp16_cfg is not None:
wrap_fp16_model(model)
load_checkpoint(model, args.checkpoint, map_location='cpu')
if args.fuse_conv_bn:
model = fuse_conv_bn(model)
if not distributed:
model = MMDataParallel(model, device_ids=[0])
outputs = single_gpu_test(model, data_loader)
else:
model = MMDistributedDataParallel(
model.cuda(),
device_ids=[torch.cuda.current_device()],
broadcast_buffers=False)
outputs = multi_gpu_test(model, data_loader, args.tmpdir,
args.gpu_collect)
rank, _ = get_dist_info()
if rank == 0:
if output_config.get('out', None):
out = output_config['out']
print(f'\nwriting results to {out}')
dataset.dump_results(outputs, **output_config)
if eval_config:
eval_res = dataset.evaluate(outputs, **eval_config)
for name, val in eval_res.items():
print(f'{name}: {val:.04f}')
def main():
args = parse_args()
cfg = Config.fromfile(args.config)
if args.cfg_options is not None:
cfg.merge_from_dict(args.cfg_options)
# import modules from string list.
if cfg.get('custom_imports', None):
from mmcv.utils import import_modules_from_strings
import_modules_from_strings(**cfg['custom_imports'])
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
# work_dir is determined in this priority: CLI > segment in file > filename
if args.work_dir is not None:
# update configs according to CLI args if args.work_dir is not None
cfg.work_dir = args.work_dir
elif cfg.get('work_dir', None) is None:
# use config filename as default work_dir if cfg.work_dir is None
cfg.work_dir = osp.join('./work_dirs',
osp.splitext(osp.basename(args.config))[0])
if args.resume_from is not None:
cfg.resume_from = args.resume_from
if args.gpu_ids is not None:
cfg.gpu_ids = args.gpu_ids
else:
cfg.gpu_ids = range(1) if args.gpus is None else range(args.gpus)
if args.autoscale_lr:
# apply the linear scaling rule (https://arxiv.org/abs/1706.02677)
cfg.optimizer['lr'] = cfg.optimizer['lr'] * len(cfg.gpu_ids) / 8
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# re-set gpu_ids with distributed training mode
_, world_size = get_dist_info()
cfg.gpu_ids = range(world_size)
# create work_dir
mmcv.mkdir_or_exist(osp.abspath(cfg.work_dir))
# dump config
cfg.dump(osp.join(cfg.work_dir, osp.basename(args.config)))
# init the logger before other steps
timestamp = time.strftime('%Y%m%d_%H%M%S', time.localtime())
log_file = osp.join(cfg.work_dir, f'{timestamp}.log')
# specify logger name, if we still use 'mmdet', the output info will be
# filtered and won't be saved in the log_file
# TODO: ugly workaround to judge whether we are training det or seg model
if cfg.model.type in ['EncoderDecoder3D']:
logger_name = 'mmseg'
else:
logger_name = 'mmdet'
logger = get_root_logger(log_file=log_file,
log_level=cfg.log_level,
name=logger_name)
# init the meta dict to record some important information such as
# environment info and seed, which will be logged
meta = dict()
# log env info
env_info_dict = collect_env()
env_info = '\n'.join([(f'{k}: {v}') for k, v in env_info_dict.items()])
dash_line = '-' * 60 + '\n'
logger.info('Environment info:\n' + dash_line + env_info + '\n' +
dash_line)
meta['env_info'] = env_info
meta['config'] = cfg.pretty_text
# log some basic info
logger.info(f'Distributed training: {distributed}')
logger.info(f'Config:\n{cfg.pretty_text}')
# set random seeds
if args.seed is not None:
logger.info(f'Set random seed to {args.seed}, '
f'deterministic: {args.deterministic}')
set_random_seed(args.seed, deterministic=args.deterministic)
cfg.seed = args.seed
meta['seed'] = args.seed
meta['exp_name'] = osp.basename(args.config)
model = build_model(cfg.model,
train_cfg=cfg.get('train_cfg'),
test_cfg=cfg.get('test_cfg'))
logger.info(f'Model:\n{model}')
datasets = [build_dataset(cfg.data.train)]
if len(cfg.workflow) == 2:
val_dataset = copy.deepcopy(cfg.data.val)
# in case we use a dataset wrapper
if 'dataset' in cfg.data.train:
val_dataset.pipeline = cfg.data.train.dataset.pipeline
else:
val_dataset.pipeline = cfg.data.train.pipeline
# set test_mode=False here in deep copied config
# which do not affect AP/AR calculation later
# refer to https://mmdetection3d.readthedocs.io/en/latest/tutorials/customize_runtime.html#customize-workflow # noqa
val_dataset.test_mode = False
datasets.append(build_dataset(val_dataset))
if cfg.checkpoint_config is not None:
# save mmdet version, config file content and class names in
# checkpoints as meta data
cfg.checkpoint_config.meta = dict(
mmdet_version=mmdet_version,
mmseg_version=mmseg_version,
mmdet3d_version=mmdet3d_version,
config=cfg.pretty_text,
CLASSES=datasets[0].CLASSES,
PALETTE=datasets[0].PALETTE # for segmentors
if hasattr(datasets[0], 'PALETTE') else None)
# add an attribute for visualization convenience
model.CLASSES = datasets[0].CLASSES
train_model(model,
datasets,
cfg,
distributed=distributed,
validate=(not args.no_validate),
timestamp=timestamp,
meta=meta)
def main():
args = parse_args()
assert args.out or args.eval or args.format_only or args.show, \
('Please specify at least one operation (save/eval/format/show the '
'results) with the argument "--out", "--eval", "--format_only" '
'or "--show"')
if args.eval and args.format_only:
raise ValueError('--eval and --format_only cannot be both specified')
if args.out is not None and not args.out.endswith(('.pkl', '.pickle')):
raise ValueError('The output file must be a pkl file.')
cfg = mmcv.Config.fromfile(args.config)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.model.pretrained = None
cfg.data.test.test_mode = True
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# build the dataloader
# TODO: support multiple images per gpu (only minor changes are needed)
dataset = build_dataset(cfg.data.test)
data_loader = build_dataloader(dataset,
samples_per_gpu=1,
workers_per_gpu=cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False)
# build the model and load checkpoint
model = build_detector(cfg.model, train_cfg=None, test_cfg=cfg.test_cfg)
fp16_cfg = cfg.get('fp16', None)
if fp16_cfg is not None:
wrap_fp16_model(model)
while not osp.isfile(args.checkpoint):
print('Waiting for {} to exist...'.format(args.checkpoint))
time.sleep(60)
checkpoint = load_checkpoint(model, args.checkpoint, map_location='cpu')
# old versions did not save class info in checkpoints, this walkaround is
# for backward compatibility
if 'CLASSES' in checkpoint['meta']:
model.CLASSES = checkpoint['meta']['CLASSES']
else:
model.CLASSES = dataset.CLASSES
if not distributed:
model = MMDataParallel(model, device_ids=[0])
outputs = single_gpu_test(model, data_loader, args.show)
else:
model = MMDistributedDataParallel(
model.cuda(),
device_ids=[torch.cuda.current_device()],
broadcast_buffers=False)
outputs = multi_gpu_test(model, data_loader, args.tmpdir,
args.gpu_collect)
rank, _ = get_dist_info()
if rank == 0:
if args.out:
print('\nwriting results to {}'.format(args.out))
mmcv.dump(outputs, args.out)
kwargs = {} if args.options is None else args.options
if args.format_only:
dataset.format_results(outputs, **kwargs)
if args.eval:
dataset.evaluate(outputs, args.eval, **kwargs)
def main():
args = parse_args()
if args.tensorrt and args.onnx:
raise ValueError(
'Cannot set onnx mode and tensorrt mode at the same time.')
cfg = Config.fromfile(args.config)
cfg.merge_from_dict(args.cfg_options)
# Load output_config from cfg
output_config = cfg.get('output_config', {})
if args.out:
# Overwrite output_config from args.out
output_config = Config._merge_a_into_b(dict(out=args.out),
output_config)
# Load eval_config from cfg
eval_config = cfg.get('eval_config', {})
if args.eval:
# Overwrite eval_config from args.eval
eval_config = Config._merge_a_into_b(dict(metrics=args.eval),
eval_config)
if args.eval_options:
# Add options from args.eval_options
eval_config = Config._merge_a_into_b(args.eval_options, eval_config)
assert output_config or eval_config, \
('Please specify at least one operation (save or eval the '
'results) with the argument "--out" or "--eval"')
dataset_type = cfg.data.test.type
if output_config.get('out', None):
if 'output_format' in output_config:
# ugly workround to make recognition and localization the same
warnings.warn(
'Skip checking `output_format` in localization task.')
else:
out = output_config['out']
# make sure the dirname of the output path exists
mmcv.mkdir_or_exist(osp.dirname(out))
_, suffix = osp.splitext(out)
if dataset_type == 'AVADataset':
assert suffix[1:] == 'csv', ('For AVADataset, the format of '
'the output file should be csv')
else:
assert suffix[1:] in file_handlers, (
'The format of the output '
'file should be json, pickle or yaml')
# set cudnn benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.data.test.test_mode = True
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# The flag is used to register module's hooks
cfg.setdefault('module_hooks', [])
# build the dataloader
dataset = build_dataset(cfg.data.test, dict(test_mode=True))
dataloader_setting = dict(videos_per_gpu=cfg.data.get('videos_per_gpu', 1),
workers_per_gpu=cfg.data.get(
'workers_per_gpu', 1),
dist=distributed,
shuffle=False)
dataloader_setting = dict(dataloader_setting,
**cfg.data.get('test_dataloader', {}))
data_loader = build_dataloader(dataset, **dataloader_setting)
if args.tensorrt:
outputs = inference_tensorrt(args.checkpoint, distributed, data_loader,
dataloader_setting['videos_per_gpu'])
elif args.onnx:
outputs = inference_onnx(args.checkpoint, distributed, data_loader,
dataloader_setting['videos_per_gpu'])
else:
outputs = inference_pytorch(args, cfg, distributed, data_loader)
rank, _ = get_dist_info()
if rank == 0:
if output_config.get('out', None):
out = output_config['out']
print(f'\nwriting results to {out}')
dataset.dump_results(outputs, **output_config)
if eval_config:
eval_res = dataset.evaluate(outputs, **eval_config)
for name, val in eval_res.items():
print(f'{name}: {val:.04f}')
def main():
args = parse_args()
if args.out is not None and not args.out.endswith(('.pkl', '.pickle')):
raise ValueError('The output file must be a pkl file.')
cfg = mmcv.Config.fromfile(args.config)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.model.pretrained = None
cfg.data.test.test_mode = True
if args.ann_file is not None:
cfg.data.test.ann_file = args.ann_file
if args.img_prefix is not None:
cfg.data.test.img_prefix = args.img_prefix
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# build the dataloader
# TODO: support multiple images per gpu (only minor changes are needed)
dataset = get_dataset(cfg.data.test)
data_loader = build_dataloader(dataset,
imgs_per_gpu=1,
workers_per_gpu=cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False)
# build the model and load checkpoint
models = []
print(f"checkpoints: {args.checkpoint}")
for checkpoint in args.checkpoint:
model = build_detector(cfg.model,
train_cfg=None,
test_cfg=cfg.test_cfg)
load_checkpoint(model, checkpoint, map_location='cpu')
models.append(model)
model = EnsembleHTC(models)
if not distributed:
model = MMDataParallel(model, device_ids=[0])
outputs = single_gpu_test(model, data_loader, args.show)
else:
model = MMDistributedDataParallel(model.cuda())
outputs = multi_gpu_test(model, data_loader, args.tmpdir)
rank, _ = get_dist_info()
if args.out and rank == 0:
print('\nwriting results to {}'.format(args.out))
mmcv.dump(outputs, args.out)
eval_types = args.eval
if eval_types:
print('Starting evaluate {}'.format(' and '.join(eval_types)))
if eval_types == ['proposal_fast']:
result_file = args.out
coco_eval(result_file, eval_types, dataset.coco)
else:
if not isinstance(outputs[0], dict):
result_file = args.out + '.json'
results2json(dataset, outputs, result_file)
coco_eval(result_file, eval_types, dataset.coco)
else:
for name in outputs[0]:
print('\nEvaluating {}'.format(name))
outputs_ = [out[name] for out in outputs]
result_file = args.out + '.{}.json'.format(name)
results2json(dataset, outputs_, result_file)
coco_eval(result_file, eval_types, dataset.coco)
def main():
args = parse_args()
assert args.out or args.eval or args.format_only or args.show \
or args.show_dir, \
('Please specify at least one operation (save/eval/format/show the '
'results / save the results) with the argument "--out", "--eval"'
', "--format-only", "--show" or "--show-dir"')
if args.eval and args.format_only:
raise ValueError('--eval and --format_only cannot be both specified')
if args.out is not None and not args.out.endswith(('.pkl', '.pickle')):
raise ValueError('The output file must be a pkl file.')
assert len(args.configs) == len(args.checkpoints)
config_list = list()
for config_path in args.configs:
cfg = Config.fromfile(config_path)
if args.cfg_options is not None:
cfg.merge_from_dict(args.cfg_options)
# import modules from string list.
if cfg.get('custom_imports', None):
from mmcv.utils import import_modules_from_strings
import_modules_from_strings(**cfg['custom_imports'])
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.model.pretrained = None
if cfg.model.get('neck'):
if isinstance(cfg.model.neck, list):
for neck_cfg in cfg.model.neck:
if neck_cfg.get('rfp_backbone'):
if neck_cfg.rfp_backbone.get('pretrained'):
neck_cfg.rfp_backbone.pretrained = None
elif cfg.model.neck.get('rfp_backbone'):
if cfg.model.neck.rfp_backbone.get('pretrained'):
cfg.model.neck.rfp_backbone.pretrained = None
if args.data_phase != 'test':
cfg.data.test = cfg.data.get(args.data_phase)
# in case the test dataset is concatenated
if isinstance(cfg.data.test, dict):
cfg.data.test.test_mode = True
elif isinstance(cfg.data.test, list):
for ds_cfg in cfg.data.test:
ds_cfg.test_mode = True
config_list.append(cfg)
primary_cfg = config_list[0]
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **primary_cfg.dist_params)
# build the dataloader
samples_per_gpu = primary_cfg.data.test.pop('samples_per_gpu', 1)
if samples_per_gpu > 1:
# Replace 'ImageToTensor' to 'DefaultFormatBundle'
primary_cfg.data.test.pipeline = replace_ImageToTensor(
primary_cfg.data.test.pipeline)
dataset = build_dataset(primary_cfg.data.test)
data_loader = build_dataloader(
dataset,
samples_per_gpu=samples_per_gpu,
workers_per_gpu=primary_cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False)
models = list()
for i, cfg in enumerate(config_list):
# build the model and load checkpoint
model = build_detector(cfg.model,
train_cfg=None,
test_cfg=cfg.test_cfg)
fp16_cfg = cfg.get('fp16', None)
if fp16_cfg is not None:
wrap_fp16_model(model)
checkpoint = load_checkpoint(model,
args.checkpoints[i],
map_location='cpu')
if args.fuse_conv_bn:
model = fuse_conv_bn(model)
# old versions did not save class info in checkpoints, this walkaround is
# for backward compatibility
if 'CLASSES' in checkpoint['meta']:
model.CLASSES = checkpoint['meta']['CLASSES']
else:
model.CLASSES = dataset.CLASSES
models.append(model)
model = HybridTaskCascadeEnsemble2(models)
if not distributed:
model = MMDataParallel(model, device_ids=[0])
os.makedirs(args.show_dir, exist_ok=True)
outputs = single_gpu_test(model, data_loader, args.show, args.show_dir,
args.show_score_thr)
else:
model = MMDistributedDataParallel(
model.cuda(),
device_ids=[torch.cuda.current_device()],
broadcast_buffers=False)
outputs = multi_gpu_test(model, data_loader, args.tmpdir,
args.gpu_collect)
rank, _ = get_dist_info()
if rank == 0:
if args.out:
print(f'\nwriting results to {args.out}')
os.makedirs(os.path.dirname(args.out), exist_ok=True)
mmcv.dump(outputs, args.out)
kwargs = {} if args.eval_options is None else args.eval_options
if args.format_only:
dataset.format_results(outputs, **kwargs)
if args.eval:
eval_kwargs = cfg.get('evaluation', {}).copy()
# hard-code way to remove EvalHook args
for key in ['interval', 'tmpdir', 'start', 'gpu_collect']:
eval_kwargs.pop(key, None)
eval_kwargs.update(dict(metric=args.eval, **kwargs))
print(dataset.evaluate(outputs, **eval_kwargs))
def main():
args = parse_args()
cfg = mmcv.Config.fromfile(args.config)
# Load output_config from cfg
output_config = cfg.get('output_config', {})
# Overwrite output_config from args.out
output_config = merge_configs(output_config, dict(out=args.out))
# Load eval_config from cfg
eval_config = cfg.get('eval_config', {})
# Overwrite eval_config from args.eval
eval_config = merge_configs(eval_config, dict(metrics=args.eval))
# Add options from args.option
eval_config = merge_configs(eval_config, args.options)
assert output_config or eval_config, \
('Please specify at least one operation (save or eval the '
'results) with the argument "--out" or "--eval"')
# set cudnn benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.data.test.test_mode = True
if cfg.test_cfg is None:
cfg.test_cfg = dict(average_clips=args.average_clips)
else:
cfg.test_cfg.average_clips = args.average_clips
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# build the dataloader
dataset = build_dataset(cfg.data.test, dict(test_mode=True))
data_loader = build_dataloader(dataset,
videos_per_gpu=1,
workers_per_gpu=cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False)
# map lable from txt to csv file
df = pd.read_csv('/data2/phap/datasets/dataset3_test.txt', header=None)
df.columns = ['full_name']
df['file_name'] = df['full_name'].apply(lambda x: x.rsplit(' ')[0])
df['true_label'] = df['full_name'].apply(lambda x: x.rsplit(' ')[-1])
# build the model and load checkpoint
model = build_model(cfg.model, train_cfg=None, test_cfg=cfg.test_cfg)
load_checkpoint(model, args.checkpoint, map_location='cpu')
if not distributed:
model = MMDataParallel(model, device_ids=[0])
outputs = single_gpu_test(model, data_loader)
else:
model = MMDistributedDataParallel(
model.cuda(),
device_ids=[torch.cuda.current_device()],
broadcast_buffers=False)
outputs = multi_gpu_test(model, data_loader, args.tmpdir,
args.gpu_collect)
# convert softmax output to one hot
pred_arr = []
for i in outputs:
pred = np.argmax(i)
pred_arr.append(pred)
# import output into csv
df['pred_label_orig'] = outputs
df['pred_label'] = pred_arr
# save csv file
df.to_csv('dataset3_test_pred_w_rwf_model.csv')
print('\nSuccess, csv file saved')
rank, _ = get_dist_info()
if rank == 0:
if output_config:
out = output_config['out']
print(f'\nwriting results to {out}')
dataset.dump_results(outputs, **output_config)
if eval_config:
eval_res = dataset.evaluate(outputs, **eval_config)
for name, val in eval_res.items():
print(f'{name}: {val:.04f}')
def main():
args = parse_args()
cfg = mmcv.Config.fromfile(args.config)
if args.update_config is not None:
cfg.merge_from_dict(args.update_config)
cfg = update_config(cfg, args)
cfg = propagate_root_dir(cfg, args.data_dir)
# Load output_config from cfg
output_config = cfg.get('output_config', {})
# Overwrite output_config from args.out
output_config = merge_configs(output_config, dict(out=args.out))
# Load eval_config from cfg
eval_config = cfg.get('eval_config', {})
# Overwrite eval_config from args.eval
eval_config = merge_configs(eval_config, dict(metrics=args.eval))
# Add options from args.option
eval_config = merge_configs(eval_config, args.options)
assert output_config or eval_config, \
('Please specify at least one operation (save or eval the '
'results) with the argument "--out" or "--eval"')
# init distributed env first, since logger depends on the dist info.
distributed = args.launcher != 'none'
if distributed:
init_dist(args.launcher, **cfg.dist_params)
# get rank
rank, _ = get_dist_info()
if cfg.get('seed'):
print(f'Set random seed to {cfg.seed}')
set_random_seed(cfg.seed)
# build the dataset
dataset = build_dataset(cfg.data, 'test', dict(test_mode=True))
if cfg.get('classes'):
dataset = dataset.filter(cfg.classes)
if rank == 0:
print(f'Test datasets:\n{str(dataset)}')
# build the dataloader
data_loader = build_dataloader(dataset,
videos_per_gpu=1,
workers_per_gpu=cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False)
# build the model and load checkpoint
model = build_model(cfg.model,
train_cfg=None,
test_cfg=cfg.test_cfg,
class_sizes=dataset.class_sizes,
class_maps=dataset.class_maps)
# nncf model wrapper
if is_checkpoint_nncf(args.checkpoint) and not cfg.get('nncf_config'):
# reading NNCF config from checkpoint
nncf_part = get_nncf_config_from_meta(args.checkpoint)
for k, v in nncf_part.items():
cfg[k] = v
if cfg.get('nncf_config'):
check_nncf_is_enabled()
if not is_checkpoint_nncf(args.checkpoint):
raise RuntimeError(
'Trying to make testing with NNCF compression a model snapshot that was NOT trained with NNCF'
)
cfg.load_from = args.checkpoint
cfg.resume_from = None
if torch.cuda.is_available():
model = model.cuda()
_, model = wrap_nncf_model(model, cfg, None, get_fake_input)
else:
fp16_cfg = cfg.get('fp16', None)
if fp16_cfg is not None:
wrap_fp16_model(model)
# load model weights
load_checkpoint(model,
args.checkpoint,
map_location='cpu',
force_matching=True)
if args.fuse_conv_bn:
model = fuse_conv_bn(model)
if not distributed:
model = MMDataParallel(model, device_ids=[0])
outputs = single_gpu_test(model, data_loader)
else:
model = MMDistributedDataParallel(
model.cuda(),
device_ids=[torch.cuda.current_device()],
broadcast_buffers=False)
outputs = multi_gpu_test(model, data_loader, args.tmpdir,
args.gpu_collect)
if rank == 0:
if output_config:
out = output_config['out']
print(f'\nwriting results to {out}')
dataset.dump_results(outputs, **output_config)
if eval_config:
eval_res = dataset.evaluate(outputs, **eval_config)
print('\nFinal metrics:')
for name, val in eval_res.items():
if 'invalid_info' in name:
continue
if isinstance(val, float):
print(f'{name}: {val:.04f}')
elif isinstance(val, str):
print(f'{name}:\n{val}')
else:
print(f'{name}: {val}')
invalid_info = {
name: val
for name, val in eval_res.items() if 'invalid_info' in name
}
if len(invalid_info) > 0:
assert args.out_invalid is not None and args.out_invalid != ''
if os.path.exists(args.out_invalid):
shutil.rmtree(args.out_invalid)
if not os.path.exists(args.out_invalid):
os.makedirs(args.out_invalid)
for name, invalid_record in invalid_info.items():
out_invalid_dir = os.path.join(args.out_invalid, name)
item_gen = zip(invalid_record['ids'],
invalid_record['conf'],
invalid_record['pred'])
for invalid_idx, pred_conf, pred_label in item_gen:
record_info = dataset.get_info(invalid_idx)
gt_label = record_info['label']
if 'filename' in record_info:
src_data_path = record_info['filename']
in_record_name, record_extension = os.path.basename(
src_data_path).split('.')
out_record_name = f'{in_record_name}_gt{gt_label}_pred{pred_label}_conf{pred_conf:.3f}'
trg_data_path = os.path.join(
out_invalid_dir,
f'{out_record_name}.{record_extension}')
shutil.copyfile(src_data_path, trg_data_path)
else:
src_data_path = record_info['frame_dir']
in_record_name = os.path.basename(src_data_path)
out_record_name = f'{in_record_name}_gt{gt_label}_pred{pred_label}_conf{pred_conf:.3f}'
trg_data_path = os.path.join(
out_invalid_dir, out_record_name)
os.makedirs(trg_data_path)
start_frame_id = record_info[
'clip_start'] + dataset.start_index
end_frame_id = record_info[
'clip_end'] + dataset.start_index
for frame_id in range(start_frame_id,
end_frame_id):
img_name = f'{frame_id:05}.jpg'
shutil.copyfile(
os.path.join(src_data_path, img_name),
os.path.join(trg_data_path, img_name))
def main():
args = parse_args()
assert args.out or args.eval or args.format_only or args.show \
or args.show_dir, \
('Please specify at least one operation (save/eval/format/show the '
'results / save the results) with the argument "--out", "--eval"'
', "--format-only", "--show" or "--show-dir"')
if args.eval and args.format_only:
raise ValueError('--eval and --format_only cannot be both specified')
if args.out is not None and not args.out.endswith(('.pkl', '.pickle')):
raise ValueError('The output file must be a pkl file.')
cfg = Config.fromfile(args.config)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.model.pretrained = None
cfg.data.test.test_mode = True
if args.city == 'shanghai_xian':
data_root = './data/BONAI/'
cfg.data.test.ann_file = data_root + 'coco/bonai_shanghai_xian_test.json'
cfg.data.test.img_prefix = data_root + 'test/images/'
else:
raise(RuntimeError("do not support the input city: ", len(args.city)))
if cfg.test_cfg.get('rcnn', False):
cfg.test_cfg.rcnn.nms.iou_threshold = args.nms_score
print("NMS config for testing: {}".format(cfg.test_cfg.rcnn.nms))
print("Dataset for evaluation: ", cfg.data.test.ann_file)
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# build the dataloader
# TODO: support multiple images per gpu (only minor changes are needed)
dataset = build_dataset(cfg.data.test)
data_loader = build_dataloader(
dataset,
samples_per_gpu=1,
workers_per_gpu=cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False)
# build the model and load checkpoint
model = build_detector(cfg.model, train_cfg=None, test_cfg=cfg.test_cfg)
fp16_cfg = cfg.get('fp16', None)
if fp16_cfg is not None:
wrap_fp16_model(model)
checkpoint = load_checkpoint(model, args.checkpoint, map_location='cpu')
if args.fuse_conv_bn:
model = fuse_module(model)
# old versions did not save class info in checkpoints, this walkaround is
# for backward compatibility
if 'CLASSES' in checkpoint['meta']:
model.CLASSES = checkpoint['meta']['CLASSES']
else:
model.CLASSES = dataset.CLASSES
if not distributed:
model = MMDataParallel(model, device_ids=[0])
outputs = single_gpu_test(model, data_loader, args.show, args.show_dir,
args.show_score_thr)
else:
model = MMDistributedDataParallel(
model.cuda(),
device_ids=[torch.cuda.current_device()],
broadcast_buffers=False)
outputs = multi_gpu_test(model, data_loader, args.tmpdir,
args.gpu_collect)
rank, _ = get_dist_info()
if rank == 0:
if args.out:
print(f'\nwriting results to {args.out}')
mmcv.dump(outputs, args.out)
kwargs = {} if args.options is None else args.options
if args.format_only:
dataset.format_results(outputs, **kwargs)
if args.eval:
dataset.evaluate(outputs, args.eval, **kwargs)
def main():
# args = ['./DOTA_configs/DOTA_hbb/retinanet_r50_fpn_2x_dota.py',
# '--gpus', '1',
# '--no-validate',
# '--work-dir', './results/retinanet_hbb_tv'
# ]
args = [
'./DOTA_configs/DOTA_obb/retinanet_r50_fpn_2x_dota.py', '--gpus', '1',
'--no-validate', '--work-dir',
'./results/retinanet_obb_tv_ver1_cv2_no_trick'
]
#
# args = ['./DOTA_configs/DOTA_obb/faster_rcnn_r50_fpn_1x_dota.py',
# '--gpus', '4',
# '--no-validate',
# '--work-dir', './results/faster_obb_tv_ver1_cv2_no_trick'
# ]
# args = ['./DOTA_configs/DOTA_obb/faster_rcnn_InLD_r50_fpn_2x_dota.py',
# '--gpus', '8',
# '--no-validate',
# '--work-dir', './results/faster_obb_tv_ver1_cv2_InLD'
# ]
# args = ['./DOTA_configs/DOTA_obb/s2anet_r50_fpn_1x_dota.py',
# './results/DOTA_s2anet_obb_tv/epoch_24.pth',
# '--out', './results/DOTA_s2anet_obb_tv/results.pkl',
# '--eval', 'bbox'
# ]
#
#
# args = ['./DOTA_configs/DIOR/retinanet_r50_fpn_2x.py',
# '--gpus', '2',
# '--no-validate',
# '--work-dir', './results/retina_test'
# ]
#
# args = ['./configs/detr/detr_r50_8x2_150e_coco.py',
# '--gpus', '4',
# # '--no-validate',
# '--work-dir', './results/detr_baseline'
# ]
# args = ['./DOTA_configs/General_RS_hbb/detr_r50_8x2_150e.py',
# '--gpus', '4',
# '--no-validate',
# '--work-dir', './results/DIOR_detr_full'
# ]
#
# args = ['./DOTA_configs/DOTA_obb/s2anet_r50_fpn_1x_dota.py',
# '--gpus', '1',
# '--no-validate',
# '--work-dir', './results/DOTA_s2anet_obb_tv'
# ]
# args = ['./DOTA_configs/DOTA_obb/faster_rcnn_r50_fpn_1x_dota.py',
# '--gpus', '1',
# '--no-validate',
# '--work-dir', './results/DOTA_faster_obb_tv_1GPU_cv2_no_trick'
# ]
args = [
'./DOTA_configs/DOTA_obb/faster_rcnn_RoITrans_r50_fpn_1x_dota.py',
'--gpus', '1', '--no-validate', '--work-dir',
'./results/DOTA_faster_rcnn_RoITrans_tv'
]
# args = ['./DOTA_configs/DOTA_obb/faster_rcnn_r50_fpn_1x_dota.py',
# '--gpus', '1',
# '--no-validate',
# '--work-dir', './results/DOTA_faster_obb_tv_1GPU_cv2_no_trick'
# ]
args = parse_args(args)
print(args)
cfg = Config.fromfile(args.config)
if args.options is not None:
cfg.merge_from_dict(args.options)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
# work_dir is determined in this priority: CLI > segment in file > filename
if args.work_dir is not None:
# update configs according to CLI args if args.work_dir is not None
cfg.work_dir = args.work_dir
elif cfg.get('work_dir', None) is None:
# use config filename as default work_dir if cfg.work_dir is None
cfg.work_dir = osp.join('./work_dirs',
osp.splitext(osp.basename(args.config))[0])
if args.resume_from is not None:
cfg.resume_from = args.resume_from
if args.gpu_ids is not None:
cfg.gpu_ids = args.gpu_ids
else:
cfg.gpu_ids = range(1) if args.gpus is None else range(args.gpus)
if args.autoscale_lr:
# apply the linear scaling rule (https://arxiv.org/abs/1706.02677)
cfg.optimizer['lr'] = cfg.optimizer['lr'] * len(cfg.gpu_ids) / 8
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# create work_dir
mmcv.mkdir_or_exist(osp.abspath(cfg.work_dir))
# dump config
cfg.dump(osp.join(cfg.work_dir, osp.basename(args.config)))
# init the logger before other steps
timestamp = time.strftime('%Y%m%d_%H%M%S', time.localtime())
log_file = osp.join(cfg.work_dir, f'{timestamp}.log')
logger = get_root_logger(log_file=log_file, log_level=cfg.log_level)
# init the meta dict to record some important information such as
# environment info and seed, which will be logged
meta = dict()
# log env info
env_info_dict = collect_env()
env_info = '\n'.join([(f'{k}: {v}') for k, v in env_info_dict.items()])
dash_line = '-' * 60 + '\n'
logger.info('Environment info:\n' + dash_line + env_info + '\n' +
dash_line)
meta['env_info'] = env_info
# log some basic info
logger.info(f'Distributed training: {distributed}')
logger.info(f'Config:\n{cfg.pretty_text}')
# set random seeds
if args.seed is not None:
logger.info(f'Set random seed to {args.seed}, '
f'deterministic: {args.deterministic}')
set_random_seed(args.seed, deterministic=args.deterministic)
cfg.seed = args.seed
meta['seed'] = args.seed
model = build_detector(cfg.model,
train_cfg=cfg.train_cfg,
test_cfg=cfg.test_cfg)
datasets = [build_dataset(cfg.data.train)]
print(len(datasets[0]))
if len(cfg.workflow) == 2:
val_dataset = copy.deepcopy(cfg.data.val)
val_dataset.pipeline = cfg.data.train.pipeline
datasets.append(build_dataset(val_dataset))
if cfg.checkpoint_config is not None:
# save mmdet version, config file content and class names in
# checkpoints as meta data
cfg.checkpoint_config.meta = dict(mmdet_version=__version__,
config=cfg.pretty_text,
CLASSES=datasets[0].CLASSES)
# add an attribute for visualization convenience
model.CLASSES = datasets[0].CLASSES
train_detector_ad(model,
datasets,
cfg,
distributed=False,
validate=(not args.no_validate),
timestamp=timestamp,
meta=meta)
def __init__(self, mode):
self.config = patch_config.patch_configs[mode]()
self.args = parse_args()
cfg = Config.fromfile(self.args.config)
cfg.data.samples_per_gpu = 1
if self.args.options is not None:
cfg.merge_from_dict(self.args.options)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
# work_dir is determined in this priority: CLI > segment in file > filename
if self.args.work_dir is not None:
if os.path.exists(self.args.work_dir) is False:
os.makedirs(self.args.work_dir)
if self.args.clear_work_dir:
file_list = os.listdir(self.args.work_dir)
for f in file_list:
if os.path.isdir(os.path.join(self.args.work_dir, f)):
shutil.rmtree(os.path.join(self.args.work_dir, f))
else:
os.remove(os.path.join(self.args.work_dir, f))
# update configs according to CLI args if args.work_dir is not None
cfg.work_dir = self.args.work_dir
elif cfg.get('work_dir', None) is None:
# use config filename as default work_dir if cfg.work_dir is None
cfg.work_dir = osp.join('./work_dirs',
osp.splitext(osp.basename(args.config))[0])
if self.args.resume_from is not None:
cfg.resume_from = self.args.resume_from
if self.args.gpu_ids is not None:
cfg.gpu_ids = self.args.gpu_ids
else:
cfg.gpu_ids = range(1) if self.args.gpus is None else range(
args.gpus)
if self.args.autoscale_lr:
# apply the linear scaling rule (https://arxiv.org/abs/1706.02677)
cfg.optimizer['lr'] = cfg.optimizer['lr'] * len(cfg.gpu_ids) / 8
# init distributed env first, since logger depends on the dist info.
if self.args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(self.args.launcher, **cfg.dist_params)
# create work_dir
mmcv.mkdir_or_exist(osp.abspath(cfg.work_dir))
# dump config
cfg.dump(osp.join(cfg.work_dir, osp.basename(self.args.config)))
# init the logger before other steps
timestamp = time.strftime('%Y%m%d_%H%M%S', time.localtime())
log_file = osp.join(cfg.work_dir, f'{timestamp}.log')
logger = get_root_logger(log_file=log_file, log_level=cfg.log_level)
# init the meta dict to record some important information such as
# environment info and seed, which will be logged
meta = dict()
# log env info
env_info_dict = collect_env()
env_info = '\n'.join([(f'{k}: {v}') for k, v in env_info_dict.items()])
dash_line = '-' * 60 + '\n'
logger.info('Environment info:\n' + dash_line + env_info + '\n' +
dash_line)
meta['env_info'] = env_info
# log some basic info
logger.info(f'Distributed training: {distributed}')
logger.info(f'Config:\n{cfg.pretty_text}')
# set random seeds
if self.args.seed is not None:
logger.info(f'Set random seed to {args.seed}, '
f'deterministic: {args.deterministic}')
set_random_seed(self.args.seed, deterministic=args.deterministic)
cfg.seed = self.args.seed
meta['seed'] = self.args.seed
self.model = build_detector(cfg.model,
train_cfg=cfg.train_cfg,
test_cfg=cfg.test_cfg)
self.model = MMDataParallel(self.model.cuda(cfg.gpu_ids[0]),
device_ids=cfg.gpu_ids)
# YOLOv4
# zzj
self.darknet_model = Darknet(self.config.cfgfile)
self.darknet_model.load_weights(self.config.weightfile)
self.darknet_model = self.darknet_model.eval().cuda(
) # TODO: Why eval?
self.datasets = [build_dataset(cfg.data.train)]
self.data_loaders = [
build_dataloader(
ds,
cfg.data.samples_per_gpu,
cfg.data.workers_per_gpu,
# cfg.gpus will be ignored if distributed
len(cfg.gpu_ids),
dist=distributed,
seed=cfg.seed) for ds in self.datasets
]
if len(cfg.workflow) == 2:
val_dataset = copy.deepcopy(cfg.data.val)
val_dataset.pipeline = cfg.data.train.pipeline
self.datasets.append(build_dataset(val_dataset))
if cfg.checkpoint_config is not None:
# save mmdet version, config file content and class names in
# checkpoints as meta data
cfg.checkpoint_config.meta = dict(mmdet_version=__version__,
config=cfg.pretty_text,
CLASSES=self.datasets[0].CLASSES)
# add an attribute for visualization convenience
self.model.CLASSES = self.datasets[0].CLASSES
self.patch_applier = PatchApplier().cuda()
self.patch_transformer = PatchTransformer().cuda()
self.prob_extractor = MaxProbExtractor(0, 80, self.config).cuda()
self.nps_calculator = NPSCalculator(self.config.printfile,
self.config.patch_size).cuda()
self.total_variation = TotalVariation().cuda()
self.writer = self.init_tensorboard(mode)
def main():
args = parse_args()
assert args.out or args.show or args.json_out, \
('Please specify at least one operation (save or show the results) '
'with the argument "--out" or "--show" or "--json_out"')
if args.out is not None and not args.out.endswith(('.pkl', '.pickle')):
raise ValueError('The output file must be a pkl file.')
if args.json_out is not None and args.json_out.endswith('.json'):
args.json_out = args.json_out[:-5]
cfg = mmcv.Config.fromfile(args.config)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.model.pretrained = None
cfg.data.test.test_mode = True
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# build the dataloader
# TODO: support multiple images per gpu (only minor changes are needed)
dataset = build_dataset(cfg.data.test)
data_loader = build_dataloader(dataset,
imgs_per_gpu=1,
workers_per_gpu=cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False)
# build the model and load checkpoint
model = build_detector(cfg.model, train_cfg=None, test_cfg=cfg.test_cfg)
fp16_cfg = cfg.get('fp16', None)
if fp16_cfg is not None:
wrap_fp16_model(model)
while not osp.isfile(args.checkpoint):
print('Waiting for {} to exist...'.format(args.checkpoint))
time.sleep(60)
checkpoint = load_checkpoint(model, args.checkpoint, map_location='cpu')
# old versions did not save class info in checkpoints, this walkaround is
# for backward compatibility
if 'CLASSES' in checkpoint['meta']:
model.CLASSES = checkpoint['meta']['CLASSES']
else:
model.CLASSES = dataset.CLASSES
if not distributed:
model = MMDataParallel(model, device_ids=[0])
outputs = single_gpu_test(model, data_loader)
else:
model = MMDistributedDataParallel(model.cuda())
outputs = multi_gpu_test(model, data_loader, args.tmpdir)
rank, _ = get_dist_info()
if args.out and rank == 0:
print('\nwriting results to {}'.format(args.out))
mmcv.dump(outputs, args.out)
eval_types = args.eval
if eval_types:
print('Starting evaluate {}'.format(' and '.join(eval_types)))
if eval_types == ['proposal_fast']:
result_file = args.out
coco_eval(result_file, eval_types, dataset.coco)
else:
if not isinstance(outputs[0], dict):
result_files = results2json_segm(dataset, outputs,
args.out)
coco_eval(result_files, eval_types, dataset.coco)
else:
for name in outputs[0]:
print('\nEvaluating {}'.format(name))
outputs_ = [out[name] for out in outputs]
result_file = args.out + '.{}'.format(name)
result_files = results2json(dataset, outputs_,
result_file)
coco_eval(result_files, eval_types, dataset.coco)
# Save predictions in the COCO json format
if args.json_out and rank == 0:
if not isinstance(outputs[0], dict):
results2json(dataset, outputs, args.json_out)
else:
for name in outputs[0]:
outputs_ = [out[name] for out in outputs]
result_file = args.json_out + '.{}'.format(name)
results2json(dataset, outputs_, result_file)
def main():
args = parse_args()
cfg = Config.fromfile(args.config)
if args.cfg_options is not None:
cfg.merge_from_dict(args.cfg_options)
# import modules from plguin/xx, registry will be updated
if hasattr(cfg, 'plugin') & cfg.plugin:
import importlib
if hasattr(cfg, 'plugin_dir'):
plugin_dir = cfg.plugin_dir
_module_dir = os.path.dirname(plugin_dir)
_module_dir = _module_dir.split('/')
_module_path = _module_dir[0]
for m in _module_dir[1:]:
_module_path = _module_path + '.' + m
print(_module_path)
plg_lib = importlib.import_module(_module_path)
else:
# import dir is the dirpath for the config file
_module_dir = os.path.dirname(args.config)
_module_dir = _module_dir.split('/')
_module_path = _module_dir[0]
for m in _module_dir[1:]:
_module_path = _module_path + '.' + m
print(_module_path)
plg_lib = importlib.import_module(_module_path)
# import modules from string list.
# if cfg.get('custom_imports', None):
# from mmcv.utils import import_modules_from_strings
# import_modules_from_strings(**cfg['custom_imports'])
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.model.pretrained = None
# in case the test dataset is concatenated
samples_per_gpu = 1
if isinstance(cfg.data.test, dict):
cfg.data.test.test_mode = True
samples_per_gpu = cfg.data.test.pop('samples_per_gpu', 1)
if samples_per_gpu > 1:
# Replace 'ImageToTensor' to 'DefaultFormatBundle'
cfg.data.test.pipeline = replace_ImageToTensor(
cfg.data.test.pipeline)
elif isinstance(cfg.data.test, list):
for ds_cfg in cfg.data.test:
ds_cfg.test_mode = True
samples_per_gpu = max(
[ds_cfg.pop('samples_per_gpu', 1) for ds_cfg in cfg.data.test])
if samples_per_gpu > 1:
for ds_cfg in cfg.data.test:
ds_cfg.pipeline = replace_ImageToTensor(ds_cfg.pipeline)
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# set random seeds
if args.seed is not None:
set_random_seed(args.seed, deterministic=args.deterministic)
# build the dataloader
dataset = build_dataset(cfg.data.test)
data_loader = build_dataloader(dataset,
samples_per_gpu=samples_per_gpu,
workers_per_gpu=cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False)
if not os.path.exists(args.out_dir):
os.mkdir(args.out_dir)
# build the model and load checkpoint
cfg.model.train_cfg = None
model = build_detector(cfg.model, test_cfg=cfg.get('test_cfg'))
#from IPython import embed
#embed()
fp16_cfg = cfg.get('fp16', None)
if fp16_cfg is not None:
wrap_fp16_model(model)
#if args.checkpoint is not None:
# checkpoint = load_checkpoint(model, args.checkpoint, map_location='cpu')
if args.fuse_conv_bn:
model = fuse_conv_bn(model)
if not distributed:
model = MMDataParallel(model, device_ids=[0])
else:
model = MMDistributedDataParallel(
model.cuda(),
device_ids=[torch.cuda.current_device()],
broadcast_buffers=False)
model.eval()
meta_json = {}
print('len of data loader: ', len(data_loader))
for i, data in tqdm(enumerate(data_loader)):
with torch.no_grad():
data = scatter(data, [-1])[0]
for k, v in data.items():
if isinstance(v, torch.Tensor):
data[k] = v.cuda()
key_img_path = data['img_metas'][0]['filename']
key_img_name = os.path.join(*key_img_path.split('/')[2:])
key_img_filename = key_img_path.split('/')[-1]
save_path = os.path.join(args.out_dir, key_img_filename)
outputs = model.module.preprocess_forward(data)
outputs = outputs.detach().cpu().numpy()
np.save(save_path, outputs)
meta_json[key_img_name] = save_path + '.npy'
with open(
os.path.join(args.out_dir,
'sf_inp_val_meta_{}.json'.format(args.local_rank)),
'w') as f:
json.dump(meta_json, f)
def main():
args = parse_args()
cfg = Config.fromfile(args.config)
cfg.merge_from_dict(args.cfg_options)
if cfg.model['test_cfg'] is None:
cfg.model['test_cfg'] = dict(feature_extraction=True)
else:
cfg.model['test_cfg']['feature_extraction'] = True
# Load output_config from cfg
output_config = cfg.get('output_config', {})
if args.out:
# Overwrite output_config from args.out
output_config = Config._merge_a_into_b(dict(out=args.out),
output_config)
assert output_config, 'Please specify output filename with --out.'
dataset_type = cfg.data.test.type
if output_config.get('out', None):
if 'output_format' in output_config:
# ugly workround to make recognition and localization the same
warnings.warn(
'Skip checking `output_format` in localization task.')
else:
out = output_config['out']
# make sure the dirname of the output path exists
mmcv.mkdir_or_exist(osp.dirname(out))
_, suffix = osp.splitext(out)
assert dataset_type == 'VideoDataset'
assert suffix[1:] in file_handlers, (
'The format of the output '
'file should be json, pickle or yaml')
# set cudnn benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.data.test.test_mode = True
cfg.data.test.data_prefix = args.video_root
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
rank, _ = get_dist_info()
size = 256
fname_tensor = torch.zeros(size, dtype=torch.uint8).cuda()
if rank == 0:
videos = open(args.video_list).readlines()
videos = [x.strip() for x in videos]
timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')
fake_anno = f'fake_anno_{timestamp}.txt'
with open(fake_anno, 'w') as fout:
lines = [x + ' 0' for x in videos]
fout.write('\n'.join(lines))
fname_tensor = text2tensor(fake_anno, size).cuda()
if distributed:
dist.broadcast(fname_tensor.cuda(), src=0)
fname = tensor2text(fname_tensor)
cfg.data.test.ann_file = fname
# The flag is used to register module's hooks
cfg.setdefault('module_hooks', [])
# build the dataloader
dataset = build_dataset(cfg.data.test, dict(test_mode=True))
dataloader_setting = dict(videos_per_gpu=cfg.data.get('videos_per_gpu', 1),
workers_per_gpu=cfg.data.get(
'workers_per_gpu', 1),
dist=distributed,
shuffle=False)
dataloader_setting = dict(dataloader_setting,
**cfg.data.get('test_dataloader', {}))
data_loader = build_dataloader(dataset, **dataloader_setting)
outputs = inference_pytorch(args, cfg, distributed, data_loader)
if rank == 0:
if output_config.get('out', None):
out = output_config['out']
print(f'\nwriting results to {out}')
dataset.dump_results(outputs, **output_config)
# remove the temporary file
os.remove(fake_anno)
def main():
args = parse_args()
cfg = Config.fromfile(args.config)
cfg.merge_from_dict(args.cfg_options)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
# work_dir is determined in this priority:
# CLI > config file > default (base filename)
if args.work_dir is not None:
# update configs according to CLI args if args.work_dir is not None
cfg.work_dir = args.work_dir
elif cfg.get('work_dir', None) is None:
# use config filename as default work_dir if cfg.work_dir is None
cfg.work_dir = osp.join('./work_dirs',
osp.splitext(osp.basename(args.config))[0])
if args.resume_from is not None:
cfg.resume_from = args.resume_from
if args.gpu_ids is not None:
cfg.gpu_ids = args.gpu_ids
else:
cfg.gpu_ids = range(1) if args.gpus is None else range(args.gpus)
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# The flag is used to determine whether it is omnisource training
cfg.setdefault('omnisource', False)
# The flag is used to register module's hooks
cfg.setdefault('module_hooks', [])
# create work_dir
mmcv.mkdir_or_exist(osp.abspath(cfg.work_dir))
# dump config
cfg.dump(osp.join(cfg.work_dir, osp.basename(args.config)))
# init logger before other steps
timestamp = time.strftime('%Y%m%d_%H%M%S', time.localtime())
log_file = osp.join(cfg.work_dir, f'{timestamp}.log')
logger = get_root_logger(log_file=log_file, log_level=cfg.log_level)
# init the meta dict to record some important information such as
# environment info and seed, which will be logged
meta = dict()
# log env info
env_info_dict = collect_env()
env_info = '\n'.join([f'{k}: {v}' for k, v in env_info_dict.items()])
dash_line = '-' * 60 + '\n'
logger.info('Environment info:\n' + dash_line + env_info + '\n' +
dash_line)
meta['env_info'] = env_info
# log some basic info
logger.info(f'Distributed training: {distributed}')
logger.info(f'Config: {cfg.pretty_text}')
# set random seeds
if args.seed is not None:
logger.info(f'Set random seed to {args.seed}, '
f'deterministic: {args.deterministic}')
set_random_seed(args.seed, deterministic=args.deterministic)
cfg.seed = args.seed
meta['seed'] = args.seed
meta['config_name'] = osp.basename(args.config)
meta['work_dir'] = osp.basename(cfg.work_dir.rstrip('/\\'))
model = build_model(
cfg.model,
train_cfg=cfg.get('train_cfg'),
test_cfg=cfg.get('test_cfg'))
if len(cfg.module_hooks) > 0:
register_module_hooks(model, cfg.module_hooks)
if cfg.omnisource:
# If omnisource flag is set, cfg.data.train should be a list
assert type(cfg.data.train) is list
datasets = [build_dataset(dataset) for dataset in cfg.data.train]
else:
datasets = [build_dataset(cfg.data.train)]
if len(cfg.workflow) == 2:
# For simplicity, omnisource is not compatiable with val workflow,
# we recommend you to use `--validate`
assert not cfg.omnisource
if args.validate:
warnings.warn('val workflow is duplicated with `--validate`, '
'it is recommended to use `--validate`. see '
'https://github.com/open-mmlab/mmaction2/pull/123')
val_dataset = copy.deepcopy(cfg.data.val)
datasets.append(build_dataset(val_dataset))
if cfg.checkpoint_config is not None:
# save mmaction version, config file content and class names in
# checkpoints as meta data
cfg.checkpoint_config.meta = dict(
mmaction_version=__version__ + get_git_hash(digits=7),
config=cfg.text)
test_option = dict(test_last=args.test_last, test_best=args.test_best)
train_model(
model,
datasets,
cfg,
distributed=distributed,
validate=args.validate,
test=test_option,
timestamp=timestamp,
meta=meta)
def main():
# options
parser = argparse.ArgumentParser()
parser.add_argument('-opt',
type=str,
required=True,
help='Path to option YAML file.')
parser.add_argument('--launcher',
choices=['none', 'pytorch', 'slurm'],
default='none',
help='job launcher')
parser.add_argument('--local_rank', type=int, default=0)
args = parser.parse_args()
opt = parse(args.opt, is_train=False)
# distributed testing settings
if args.launcher == 'none': # non-distributed testing
opt['dist'] = False
print('Disable distributed testing.', flush=True)
else:
opt['dist'] = True
if args.launcher == 'slurm' and 'dist_params' in opt:
init_dist(args.launcher, **opt['dist_params'])
else:
init_dist(args.launcher)
rank, world_size = get_dist_info()
opt['rank'] = rank
opt['world_size'] = world_size
make_exp_dirs(opt)
log_file = osp.join(opt['path']['log'],
f"test_{opt['name']}_{get_time_str()}.log")
logger = get_root_logger(logger_name='basicsr',
log_level=logging.INFO,
log_file=log_file)
logger.info(get_env_info())
logger.info(dict2str(opt))
# random seed
seed = opt['manual_seed']
if seed is None:
seed = random.randint(1, 10000)
opt['manual_seed'] = seed
logger.info(f'Random seed: {seed}')
set_random_seed(seed + rank)
torch.backends.cudnn.benchmark = True
# torch.backends.cudnn.deterministic = True
# create test dataset and dataloader
test_loaders = []
for phase, dataset_opt in sorted(opt['datasets'].items()):
test_set = create_dataset(dataset_opt)
test_loader = create_dataloader(test_set,
dataset_opt,
num_gpu=opt['num_gpu'],
dist=opt['dist'],
sampler=None,
seed=seed)
logger.info(
f"Number of test images in {dataset_opt['name']}: {len(test_set)}")
test_loaders.append(test_loader)
# create model
model = create_model(opt)
for test_loader in test_loaders:
test_set_name = test_loader.dataset.opt['name']
logger.info(f'Testing {test_set_name}...')
model.validation(test_loader,
current_iter=opt['name'],
tb_logger=None,
save_img=True)
dummy_input = torch.randn(1, 7, 3, 144, 176, device='cuda')
input_names = ['input']
output_names = ['output']
torch.onnx.export(
model.net_g.module,
dummy_input,
"trt_onnx/edvr.onnx",
verbose=True,
opset_version=11,
operator_export_type=torch.onnx.OperatorExportTypes.ONNX_FALLTHROUGH,
input_names=input_names,
output_names=output_names)
def main():
args = parse_args()
cfg = Config.fromfile(args.config)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
# update configs according to CLI args
if args.work_dir is not None:
cfg.work_dir = args.work_dir
if args.resume_from is not None:
cfg.resume_from = args.resume_from
cfg.gpus = args.gpus
if args.autoscale_lr:
# apply the linear scaling rule (https://arxiv.org/abs/1706.02677)
cfg.optimizer['lr'] = cfg.optimizer['lr'] * cfg.gpus / 8
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# create work_dir
mmcv.mkdir_or_exist(osp.abspath(cfg.work_dir))
# init the logger before other steps
timestamp = time.strftime('%Y%m%d_%H%M%S', time.localtime())
log_file = osp.join(cfg.work_dir, '{}.log'.format(timestamp))
logger = get_root_logger(log_file=log_file, log_level=cfg.log_level)
# init the meta dict to record some important information such as
# environment info and seed, which will be logged
meta = dict()
# log env info
env_info_dict = collect_env()
env_info = '\n'.join([('{}: {}'.format(k, v))
for k, v in env_info_dict.items()])
dash_line = '-' * 60 + '\n'
logger.info('Environment info:\n' + dash_line + env_info + '\n' +
dash_line)
meta['env_info'] = env_info
# log some basic info
logger.info('Distributed training: {}'.format(distributed))
logger.info('Config:\n{}'.format(cfg.text))
# set random seeds
if args.seed is not None:
logger.info('Set random seed to {}, deterministic: {}'.format(
args.seed, args.deterministic))
set_random_seed(args.seed, deterministic=args.deterministic)
cfg.seed = args.seed
meta['seed'] = args.seed
model = build_detector(cfg.model,
train_cfg=cfg.train_cfg,
test_cfg=cfg.test_cfg)
datasets = [build_dataset(cfg.data.train)]
if len(cfg.workflow) == 2:
val_dataset = copy.deepcopy(cfg.data.val)
val_dataset.pipeline = cfg.data.train.pipeline
datasets.append(build_dataset(val_dataset))
if cfg.checkpoint_config is not None:
# save mmdet version, config file content and class names in
# checkpoints as meta data
cfg.checkpoint_config.meta = dict(mmdet_version=__version__,
config=cfg.text,
CLASSES=datasets[0].CLASSES)
# add an attribute for visualization convenience
model.CLASSES = datasets[0].CLASSES
train_detector(model,
datasets,
cfg,
distributed=distributed,
validate=args.validate,
timestamp=timestamp,
meta=meta)
def main():
# options
parser = argparse.ArgumentParser()
parser.add_argument('-opt',
type=str,
required=True,
help='Path to option YAML file.')
parser.add_argument('--launcher',
choices=['none', 'pytorch', 'slurm'],
default='none',
help='job launcher')
parser.add_argument('--local_rank', type=int, default=0)
args = parser.parse_args()
opt = parse(args.opt, is_train=True)
# distributed training settings
if args.launcher == 'none': # non-distributed training
opt['dist'] = False
print('Disable distributed training.')
else:
opt['dist'] = True
if args.launcher == 'slurm' and 'dist_params' in opt:
init_dist(args.launcher, **opt['dist_params'])
else:
init_dist(args.launcher)
rank, world_size = get_dist_info()
opt['rank'] = rank
opt['world_size'] = world_size
# load resume states if exists
if opt['path'].get('resume_state'):
device_id = torch.cuda.current_device()
resume_state = torch.load(
opt['path']['resume_state'],
map_location=lambda storage, loc: storage.cuda(device_id))
else:
resume_state = None
# mkdir and loggers
if resume_state is None:
make_exp_dirs(opt)
log_file = './first_xavier_L_40.log'
logger = get_root_logger(logger_name='basicsr',
log_level=logging.INFO,
log_file=log_file)
logger.info(get_env_info())
logger.info(dict2str(opt))
# initialize tensorboard logger and wandb logger
tb_logger = None
if opt['logger'].get('use_tb_logger') and 'debug' not in opt['name']:
log_dir = './tb_logger/' + opt['name']
if resume_state is None and opt['rank'] == 0:
mkdir_and_rename(log_dir)
tb_logger = init_tb_logger(log_dir=log_dir)
if (opt['logger'].get('wandb')
is not None) and (opt['logger']['wandb'].get('project')
is not None) and ('debug' not in opt['name']):
assert opt['logger'].get('use_tb_logger') is True, (
'should turn on tensorboard when using wandb')
init_wandb_logger(opt)
# random seed
print('1')
seed = opt['manual_seed']
# if seed is None:
# seed = random.randint(1, 10000)
# opt['manual_seed'] = seed
set_random_seed(seed + rank)
torch.backends.cudnn.benchmark = True
# torch.backends.cudnn.deterministic = True
# create train and val dataloaders
train_loader, val_loader = None, None
for phase, dataset_opt in opt['datasets'].items():
if phase == 'train':
dataset_enlarge_ratio = dataset_opt.get('dataset_enlarge_ratio', 1)
train_set = create_dataset(dataset_opt)
train_sampler = EnlargedSampler(train_set, world_size, rank,
dataset_enlarge_ratio)
train_loader = create_dataloader(train_set,
dataset_opt,
num_gpu=opt['num_gpu'],
dist=opt['dist'],
sampler=train_sampler,
seed=seed)
num_iter_per_epoch = math.ceil(
len(train_set) * dataset_enlarge_ratio /
(dataset_opt['batch_size_per_gpu'] * opt['world_size']))
total_iters = int(opt['train']['total_iter'])
total_epochs = math.ceil(total_iters / (num_iter_per_epoch))
logger.info(
'Training statistics:'
f'\n\tNumber of train images: {len(train_set)}'
f'\n\tDataset enlarge ratio: {dataset_enlarge_ratio}'
f'\n\tBatch size per gpu: {dataset_opt["batch_size_per_gpu"]}'
f'\n\tWorld size (gpu number): {opt["world_size"]}'
f'\n\tRequire iter number per epoch: {num_iter_per_epoch}'
f'\n\tTotal epochs: {total_epochs}; iters: {total_iters}.')
elif phase == 'val':
val_set = create_dataset(dataset_opt)
val_loader = create_dataloader(val_set,
dataset_opt,
num_gpu=opt['num_gpu'],
dist=opt['dist'],
sampler=None,
seed=seed)
logger.info(
f'Number of val images/folders in {dataset_opt["name"]}: '
f'{len(val_set)}')
else:
raise ValueError(f'Dataset phase {phase} is not recognized.')
assert train_loader is not None
# create model
if resume_state:
check_resume(opt, resume_state['iter']) # modify pretrain_model paths
model = create_model(opt)
# resume training
if resume_state:
logger.info(f"Resuming training from epoch: {resume_state['epoch']}, "
f"iter: {resume_state['iter']}.")
start_epoch = resume_state['epoch']
current_iter = resume_state['iter']
model.resume_training(resume_state) # handle optimizers and schedulers
else:
start_epoch = 0
current_iter = 0
# create message logger (formatted outputs)
msg_logger = MessageLogger(opt, current_iter, tb_logger)
# dataloader prefetcher
prefetch_mode = opt['datasets']['train'].get('prefetch_mode')
if prefetch_mode is None or prefetch_mode == 'cpu':
prefetcher = CPUPrefetcher(train_loader)
elif prefetch_mode == 'cuda':
prefetcher = CUDAPrefetcher(train_loader, opt)
logger.info(f'Use {prefetch_mode} prefetch dataloader')
if opt['datasets']['train'].get('pin_memory') is not True:
raise ValueError('Please set pin_memory=True for CUDAPrefetcher.')
else:
raise ValueError(f'Wrong prefetch_mode {prefetch_mode}.'
"Supported ones are: None, 'cuda', 'cpu'.")
# training
logger.info(
f'Start training from epoch: {start_epoch}, iter: {current_iter}')
data_time, iter_time = time.time(), time.time()
start_time = time.time()
for epoch in range(start_epoch, total_epochs + 1):
train_sampler.set_epoch(epoch)
prefetcher.reset()
train_data = prefetcher.next()
while train_data is not None:
data_time = time.time() - data_time
current_iter += 1
if current_iter > total_iters:
break
# update learning rate
model.update_learning_rate(current_iter,
warmup_iter=opt['train'].get(
'warmup_iter', -1))
# training
model.feed_data(train_data)
model.optimize_parameters(current_iter)
iter_time = time.time() - iter_time
# log
if current_iter % opt['logger']['print_freq'] == 0:
log_vars = {'epoch': epoch, 'iter': current_iter}
log_vars.update({'lrs': model.get_current_learning_rate()})
log_vars.update({'time': iter_time, 'data_time': data_time})
log_vars.update(model.get_current_log())
msg_logger(log_vars)
# save models and training states
if current_iter % opt['logger']['save_checkpoint_freq'] == 0:
logger.info('Saving models and training states.')
model.save(epoch, current_iter)
# validation
if opt['val']['val_freq'] is not None and current_iter % opt[
'val']['val_freq'] == 0:
model.validation(val_loader, current_iter, tb_logger,
opt['val']['save_img'])
data_time = time.time()
iter_time = time.time()
train_data = prefetcher.next()
# end of iter
# end of epoch
consumed_time = str(
datetime.timedelta(seconds=int(time.time() - start_time)))
logger.info(f'End of training. Time consumed: {consumed_time}')
logger.info('Save the latest model.')
model.save(epoch=-1, current_iter=-1) # -1 stands for the latest
if opt['val']['val_freq'] is not None:
model.validation(val_loader, current_iter, tb_logger,
opt['val']['save_img'])
if tb_logger:
tb_logger.close()
def main():
global cfg, rank, world_size
cfg = Config.fromfile(args.config)
# Set seed
np.random.seed(cfg.seed)
cudnn.benchmark = True
torch.manual_seed(cfg.seed)
cudnn.enabled = True
torch.cuda.manual_seed(cfg.seed)
# Model
print('==> Building model..')
arch_code = eval('architecture_code.{}'.format(cfg.model))
net = models.model_entry(cfg, arch_code)
rank = 0 # for non-distributed
world_size = 1 # for non-distributed
if args.distributed:
print('==> Initializing distributed training..')
init_dist(
launcher='slurm', backend='nccl'
) # Only support slurm for now, if you would like to personalize your launcher, please refer to https://github.com/open-mmlab/mmcv/blob/master/mmcv/runner/dist_utils.py
rank, world_size = get_dist_info()
net = net.cuda()
cfg.netpara = sum(p.numel() for p in net.parameters()) / 1e6
start_epoch = 0
best_acc = 0
# Load checkpoint.
if cfg.get('resume_path', False):
print('==> Resuming from {}checkpoint {}..'.format(
('original ' if cfg.resume_path.origin_ckpt else ''),
cfg.resume_path.path))
if cfg.resume_path.origin_ckpt:
utils.load_state(cfg.resume_path.path, net, rank=rank)
else:
if args.distributed:
net = torch.nn.parallel.DistributedDataParallel(
net,
device_ids=[torch.cuda.current_device()],
output_device=torch.cuda.current_device())
utils.load_state(cfg.resume_path.path, net, rank=rank)
# Data
print('==> Preparing data..')
trainloader, testloader, train_sampler, test_sampler = dataset_entry(
cfg, args.distributed)
criterion = nn.CrossEntropyLoss()
if not args.eval_only:
cfg.attack_param.num_steps = 7
net_adv = AttackPGD(net, cfg.attack_param)
# Train params
print('==> Setting train parameters..')
train_param = cfg.train_param
epochs = train_param.epochs
init_lr = train_param.learning_rate
if train_param.get('warm_up_param', False):
warm_up_param = train_param.warm_up_param
init_lr = warm_up_param.warm_up_base_lr
epochs += warm_up_param.warm_up_epochs
if train_param.get('no_wd', False):
param_group, type2num, _, _ = utils.param_group_no_wd(net)
cfg.param_group_no_wd = type2num
optimizer = torch.optim.SGD(param_group,
lr=init_lr,
momentum=train_param.momentum,
weight_decay=train_param.weight_decay)
else:
optimizer = torch.optim.SGD(net.parameters(),
lr=init_lr,
momentum=train_param.momentum,
weight_decay=train_param.weight_decay)
scheduler = lr_scheduler.CosineLRScheduler(
optimizer, epochs, train_param.learning_rate_min, init_lr,
train_param.learning_rate,
(warm_up_param.warm_up_epochs if train_param.get(
'warm_up_param', False) else 0))
# Log
print('==> Writing log..')
if rank == 0:
cfg.save = '{}/{}-{}-{}'.format(cfg.save_path, cfg.model, cfg.dataset,
time.strftime("%Y%m%d-%H%M%S"))
utils.create_exp_dir(cfg.save)
logger = utils.create_logger('global_logger', cfg.save + '/log.txt')
logger.info('config: {}'.format(pprint.pformat(cfg)))
# Evaluation only
if args.eval_only:
assert cfg.get(
'resume_path',
False), 'Should set the resume path for the eval_only mode'
print('==> Testing on Clean Data..')
test(net, testloader, criterion)
print('==> Testing on Adversarial Data..')
test(net_adv, testloader, criterion, adv=True)
return
# Training process
for epoch in range(start_epoch, epochs):
train_sampler.set_epoch(epoch)
test_sampler.set_epoch(epoch)
scheduler.step()
if rank == 0:
logger.info('Epoch %d learning rate %e', epoch,
scheduler.get_lr()[0])
# Train for one epoch
train(net_adv, trainloader, criterion, optimizer)
# Validate for one epoch
valid_acc = test(net_adv, testloader, criterion, adv=True)
if rank == 0:
logger.info('Validation Accuracy: {}'.format(valid_acc))
is_best = valid_acc > best_acc
best_acc = max(valid_acc, best_acc)
print('==> Saving')
state = {
'epoch': epoch,
'best_acc': best_acc,
'optimizer': optimizer.state_dict(),
'state_dict': net.state_dict(),
'scheduler': scheduler
}
utils.save_checkpoint(state, is_best, os.path.join(cfg.save))
def main():
args = parse_args()
cfg = Config.fromfile(args.config)
if args.cfg_options is not None:
cfg.merge_from_dict(args.cfg_options)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
if args.gpu_ids is not None:
cfg.gpu_ids = args.gpu_ids[0:1]
warnings.warn('`--gpu-ids` is deprecated, please use `--gpu-id`. '
'Because we only support single GPU mode in '
'non-distributed testing. Use the first GPU '
'in `gpu_ids` now.')
else:
cfg.gpu_ids = [args.gpu_id]
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
rank = 0
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
rank, world_size = get_dist_info()
cfg.gpu_ids = range(world_size)
assert args.online or world_size == 1, (
'We only support online mode for distrbuted evaluation.')
dirname = os.path.dirname(args.checkpoint)
ckpt = os.path.basename(args.checkpoint)
if 'http' in args.checkpoint:
log_path = None
else:
log_name = ckpt.split('.')[0] + '_eval_log' + '.txt'
log_path = os.path.join(dirname, log_name)
logger = get_root_logger(log_file=log_path,
log_level=cfg.log_level,
file_mode='a')
logger.info('evaluation')
# set random seeds
if args.seed is not None:
if rank == 0:
mmcv.print_log(f'set random seed to {args.seed}', 'mmgen')
set_random_seed(args.seed, deterministic=args.deterministic)
# build the model and load checkpoint
model = build_model(cfg.model,
train_cfg=cfg.train_cfg,
test_cfg=cfg.test_cfg)
# sanity check for models without ema
if not model.use_ema:
args.sample_model = 'orig'
mmcv.print_log(f'Sampling model: {args.sample_model}', 'mmgen')
model.eval()
if args.eval:
if args.eval[0] == 'none':
# only sample images
metrics = []
assert args.num_samples is not None and args.num_samples > 0
else:
metrics = [
build_metric(cfg.metrics[metric]) for metric in args.eval
]
else:
metrics = [build_metric(cfg.metrics[metric]) for metric in cfg.metrics]
# check metrics for dist evaluation
if distributed and metrics:
for metric in metrics:
assert metric.name in _distributed_metrics, (
f'We only support {_distributed_metrics} for multi gpu '
f'evaluation, but receive {args.eval}.')
_ = load_checkpoint(model, args.checkpoint, map_location='cpu')
basic_table_info = dict(train_cfg=os.path.basename(cfg._filename),
ckpt=ckpt,
sample_model=args.sample_model)
if len(metrics) == 0:
basic_table_info['num_samples'] = args.num_samples
data_loader = None
else:
basic_table_info['num_samples'] = -1
# build the dataloader
if cfg.data.get('test', None) and cfg.data.test.get('imgs_root', None):
dataset = build_dataset(cfg.data.test)
elif cfg.data.get('val', None) and cfg.data.val.get('imgs_root', None):
dataset = build_dataset(cfg.data.val)
elif cfg.data.get('train', None):
# we assume that the train part should work well
dataset = build_dataset(cfg.data.train)
else:
raise RuntimeError('There is no valid dataset config to run, '
'please check your dataset configs.')
# The default loader config
loader_cfg = dict(samples_per_gpu=args.batch_size,
workers_per_gpu=cfg.data.get(
'val_workers_per_gpu', cfg.data.workers_per_gpu),
num_gpus=len(cfg.gpu_ids),
dist=distributed,
shuffle=True)
# The overall dataloader settings
loader_cfg.update({
k: v
for k, v in cfg.data.items() if k not in [
'train', 'val', 'test', 'train_dataloader', 'val_dataloader',
'test_dataloader'
]
})
# specific config for test loader
test_loader_cfg = {**loader_cfg, **cfg.data.get('test_dataloader', {})}
data_loader = build_dataloader(dataset, **test_loader_cfg)
if args.sample_cfg is None:
args.sample_cfg = dict()
if not distributed:
model = MMDataParallel(model, device_ids=[0])
else:
find_unused_parameters = cfg.get('find_unused_parameters', False)
model = MMDistributedDataParallel(
model.cuda(),
device_ids=[torch.cuda.current_device()],
broadcast_buffers=False,
find_unused_parameters=find_unused_parameters)
# online mode will not save samples
if args.online and len(metrics) > 0:
online_evaluation(model, data_loader, metrics, logger,
basic_table_info, args.batch_size, **args.sample_cfg)
else:
offline_evaluation(model, data_loader, metrics, logger,
basic_table_info, args.batch_size,
args.samples_path, **args.sample_cfg)
type=str,
help='Specify a config file path')
parser.add_argument('--launcher', default=None, type=str, help='Launcher')
parser.add_argument(
'--num_workers',
default=4,
type=int,
help='Specify the number of worker threads for data loaders')
parser.add_argument('--local_rank', type=int, default=0)
parser.add_argument('--seed',
default=0,
type=int,
help='Specify a random seed')
args = parser.parse_args()
## set up
cfgs = setup_runtime(args)
if args.launcher is None or args.launcher == 'none':
cfgs['distributed'] = False
else:
cfgs['distributed'] = True
init_dist(args.launcher, backend='nccl')
# important: use different random seed for different process
torch.manual_seed(args.seed + dist.get_rank())
print(cfgs)
trainer = Trainer(cfgs, GAN2Shape)
## run
trainer.train()
def main():
args = parse_args()
cfg = Config.fromfile(args.config)
if args.options is not None:
cfg.merge_from_dict(args.options)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
# work_dir is determined in this priority: CLI > segment in file > filename
if args.work_dir is not None:
# update configs according to CLI args if args.work_dir is not None
cfg.work_dir = args.work_dir
elif cfg.get('work_dir', None) is None:
# use config filename as default work_dir if cfg.work_dir is None
cfg.work_dir = osp.join('./work_dirs',
osp.splitext(osp.basename(args.config))[0])
if args.load_from is not None:
cfg.load_from = args.load_from
if args.resume_from is not None:
cfg.resume_from = args.resume_from
if args.gpu_ids is not None:
cfg.gpu_ids = args.gpu_ids
else:
cfg.gpu_ids = range(1) if args.gpus is None else range(args.gpus)
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# create work_dir
mmcv.mkdir_or_exist(osp.abspath(cfg.work_dir))
# dump config
cfg.dump(osp.join(cfg.work_dir, osp.basename(args.config)))
# init the logger before other steps
timestamp = time.strftime('%Y%m%d_%H%M%S', time.localtime())
log_file = osp.join(cfg.work_dir, f'{timestamp}.log')
logger = get_root_logger(log_file=log_file, log_level=cfg.log_level)
# init the meta dict to record some important information such as
# environment info and seed, which will be logged
meta = dict()
# log env info
env_info_dict = collect_env()
env_info = '\n'.join([f'{k}: {v}' for k, v in env_info_dict.items()])
dash_line = '-' * 60 + '\n'
logger.info('Environment info:\n' + dash_line + env_info + '\n' +
dash_line)
meta['env_info'] = env_info
# log some basic info
logger.info(f'Distributed training: {distributed}')
logger.info(f'Config:\n{cfg.pretty_text}')
# set random seeds
if args.seed is not None:
logger.info(f'Set random seed to {args.seed}, deterministic: '
f'{args.deterministic}')
set_random_seed(args.seed, deterministic=args.deterministic)
cfg.seed = args.seed
meta['seed'] = args.seed
meta['exp_name'] = osp.basename(args.config)
# T_config = 'configs/pspnet/pspnet_r18-d8_512x512_nophoto_40k_cityscapes.py'
# T_cfg = Config.fromfile(T_config)
# # print(T_cfg)
# Teacher = build_segmentor(T_cfg.model, train_cfg=T_cfg.train_cfg, test_cfg=T_cfg.test_cfg)
# # cfg.model.teacher = Teacher
# print('===========================================')
# # print(type(cfg))
# print(type(Teacher))
# model = build_segmentor_kd(
# cfg.model, train_cfg=cfg.train_cfg, test_cfg=cfg.test_cfg, teacher=Teacher)
model = build_segmentor(cfg.model,
train_cfg=cfg.train_cfg,
test_cfg=cfg.test_cfg)
logger.info(model)
datasets = [build_dataset(cfg.data.train)]
if len(cfg.workflow) == 2:
val_dataset = copy.deepcopy(cfg.data.val)
val_dataset.pipeline = cfg.data.train.pipeline
datasets.append(build_dataset(val_dataset))
if cfg.checkpoint_config is not None:
# save mmseg version, config file content and class names in
# checkpoints as meta data
cfg.checkpoint_config.meta = dict(
mmseg_version=f'{__version__}+{get_git_hash()[:7]}',
config=cfg.pretty_text,
CLASSES=datasets[0].CLASSES,
PALETTE=datasets[0].PALETTE)
# add an attribute for visualization convenience
model.CLASSES = datasets[0].CLASSES
train_segmentor(model,
datasets,
cfg,
distributed=distributed,
validate=(not args.no_validate),
timestamp=timestamp,
meta=meta)
def main():
args = parse_args()
cfg = mmcv.Config.fromfile(args.config)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
# update configs according to CLI args
if args.work_dir is not None:
cfg.work_dir = args.work_dir
layer_ind = [int(idx) for idx in args.layer_ind.split(',')]
cfg.model.backbone.out_indices = layer_ind
if args.checkpoint is None:
assert cfg.model.pretrained is not None, \
"Must have pretrain if no checkpoint is given."
# check memcached package exists
if importlib.util.find_spec('mc') is None:
for field in ['train', 'val', 'test']:
if hasattr(cfg.data, field):
getattr(cfg.data, field).data_source.memcached = False
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# logger
timestamp = time.strftime('%Y%m%d_%H%M%S', time.localtime())
log_file = osp.join(cfg.work_dir, 'extract_{}.log'.format(timestamp))
logger = get_root_logger(log_file=log_file, log_level=cfg.log_level)
# build the dataloader
dataset_cfg = mmcv.Config.fromfile(args.dataset_config)
dataset = build_dataset(dataset_cfg.data.extract)
data_loader = build_dataloader(
dataset,
imgs_per_gpu=dataset_cfg.data.imgs_per_gpu,
workers_per_gpu=dataset_cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False)
# build the model and load checkpoint
model = build_model(cfg.model)
if args.checkpoint is not None:
load_checkpoint(model, args.checkpoint, map_location='cpu')
if not distributed:
model = MMDataParallel(model, device_ids=[0])
else:
model = MMDistributedDataParallel(
model.cuda(),
device_ids=[torch.cuda.current_device()],
broadcast_buffers=False)
# build extraction processor
extractor = ExtractProcess(pool_type='specified',
backbone='resnet50',
layer_indices=layer_ind)
# run
outputs = extractor.extract(model, data_loader, distributed=distributed)
rank, _ = get_dist_info()
mmcv.mkdir_or_exist("{}/features/".format(args.work_dir))
if rank == 0:
for key, val in outputs.items():
split_num = len(dataset_cfg.split_name)
split_at = dataset_cfg.split_at
for ss in range(split_num):
output_file = "{}/features/{}_{}.npy".format(
args.work_dir, dataset_cfg.split_name[ss], key)
if ss == 0:
np.save(output_file, val[:split_at[0]])
elif ss == split_num - 1:
np.save(output_file, val[split_at[-1]:])
else:
np.save(output_file, val[split_at[ss - 1]:split_at[ss]])
def main():
args = parse_args()
cfg = Config.fromfile(args.config)
if args.options is not None:
cfg.merge_from_dict(args.options)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
# work_dir is determined in this priority: CLI > segment in file > filename
if args.work_dir is not None:
# update configs according to CLI args if args.work_dir is not None
cfg.work_dir = args.work_dir
elif cfg.get('work_dir', None) is None:
# use config filename as default work_dir if cfg.work_dir is None
cfg.work_dir = osp.join('./work_dirs',
osp.splitext(osp.basename(args.config))[0])
if args.resume_from is not None:
cfg.resume_from = args.resume_from
if args.gpu_ids is not None:
cfg.gpu_ids = args.gpu_ids
else:
cfg.gpu_ids = range(1) if args.gpus is None else range(args.gpus)
if args.autoscale_lr:
# apply the linear scaling rule (https://arxiv.org/abs/1706.02677)
cfg.optimizer['lr'] = cfg.optimizer['lr'] * len(cfg.gpu_ids) / 8
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# create work_dir
mmcv.mkdir_or_exist(osp.abspath(cfg.work_dir))
# init the logger before other steps
timestamp = time.strftime('%Y%m%d_%H%M%S', time.localtime())
log_file = osp.join(cfg.work_dir, f'{timestamp}.log')
logger = get_root_logger(log_file=log_file, log_level=cfg.log_level)
# add a logging filter
logging_filter = logging.Filter('mmdet')
logging_filter.filter = lambda record: record.find('mmdet') != -1
# init the meta dict to record some important information such as
# environment info and seed, which will be logged
meta = dict()
# log env info
env_info_dict = collect_env()
env_info = '\n'.join([(f'{k}: {v}') for k, v in env_info_dict.items()])
dash_line = '-' * 60 + '\n'
logger.info('Environment info:\n' + dash_line + env_info + '\n' +
dash_line)
meta['env_info'] = env_info
# log some basic info
logger.info(f'Distributed training: {distributed}')
logger.info(f'Config:\n{cfg.pretty_text}')
# set random seeds
if args.seed is not None:
logger.info(f'Set random seed to {args.seed}, '
f'deterministic: {args.deterministic}')
set_random_seed(args.seed, deterministic=args.deterministic)
cfg.seed = args.seed
meta['seed'] = args.seed
model = build_detector(cfg.model,
train_cfg=cfg.train_cfg,
test_cfg=cfg.test_cfg)
logger.info(f'Model:\n{model}')
# datasets = [build_dataset(cfg.data.train)]
datasets = [build_dataset(cfg.data.source_train)]
if len(cfg.workflow) == 2:
val_dataset = copy.deepcopy(cfg.data.val)
val_dataset.pipeline = cfg.data.train.pipeline
datasets.append(build_dataset(val_dataset))
if cfg.checkpoint_config is not None:
# save mmdet version, config file content and class names in
# checkpoints as meta data
cfg.checkpoint_config.meta = dict(mmdet_version=__version__,
config=cfg.pretty_text,
CLASSES=datasets[0].CLASSES)
# add an attribute for visualization convenience
model.CLASSES = datasets[0].CLASSES
train_source_detector(model,
datasets,
cfg,
distributed=distributed,
timestamp=timestamp,
meta=meta)
def main():
data_length = 2000
args = parse_args()
cfg = Config.fromfile(args.config)
cfg.seed = args.seed
cfg.data.samples_per_gpu = 2
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.model.pretrained = None
cfg.data.test.test_mode = True
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
if args.gpu_ids is not None:
cfg.gpu_ids = args.gpu_ids
else:
cfg.gpu_ids = range(1) if args.gpus is None else range(args.gpus)
# build the dataloader
# TODO: support multiple images per gpu (only minor changes are needed)
datasets = [build_dataset(cfg.data.train)]
data_loaders = [
build_dataloader(
ds,
cfg.data.samples_per_gpu,
cfg.data.workers_per_gpu,
# cfg.gpus will be ignored if distributed
len(cfg.gpu_ids),
dist=distributed,
seed=cfg.seed) for ds in datasets
]
data_loader = data_loaders[0]
output = []
ia = 0
for i_batch, data in tqdm(enumerate(data_loader), total=len(data_loader)):
if len(output) >= data_length:
break
imgs = data['img'].data[0]
img_metas = data['img_metas'].data[0]
img_tensor = imgs.detach()
imgs = tensor2imgs(img_tensor, **img_metas[0]['img_norm_cfg'])
for i in range(len(imgs)):
h, w, _ = img_metas[i]['img_shape']
img = imgs[i][:h, :w, :]
gt_bboxes = data['gt_bboxes'].data[0][i]
area = torch.mul((gt_bboxes[:, 2] - gt_bboxes[:, 0]),
(gt_bboxes[:, 3] - gt_bboxes[:, 1]))
if torch.min(area / (h * w)) < 0.05:
continue
if 4 <= gt_bboxes.size()[0] <= 15:
# output.append(cv2.resize(img, (1000, 1000), interpolation=cv2.INTER_CUBIC))
img_o = cv2.resize(img, (1000, 1000),
interpolation=cv2.INTER_CUBIC)
mmcv.imwrite(img_o, args.save_dir + '/' + str(ia) + '.png')
print(ia)
ia += 1
print('finish')