dataset_train, dataset_test = get_dataset(args)
time_limits = 5 * 60 * 60 # 5 days
epochs = 12
# use coco pre-trained model for custom datasets
transfer = None if ('voc' in args.dataset_name) or (
'coco' in args.dataset_name) else 'coco'
detector = task.fit(dataset_train,
num_trials=30,
epochs=epochs,
net=ag.Categorical('darknet53', 'mobilenet1.0'),
meta_arch=args.meta_arch,
lr=ag.Categorical(1e-2, 5e-3, 1e-3, 5e-4, 1e-4, 5e-5),
transfer=transfer,
data_shape=ag.Categorical(320, 416),
nthreads_per_trial=16,
ngpus_per_trial=1,
batch_size=8,
lr_decay_epoch=ag.Categorical('80,90', '85,95'),
warmup_epochs=ag.Int(1, 10),
warmup_iters=ag.Int(250, 1000),
wd=ag.Categorical(1e-4, 5e-4, 2.5e-4),
syncbn=ag.Bool(),
label_smooth=ag.Bool(),
time_limits=time_limits,
dist_ip_addrs=[])
test_map = detector.evaluate(dataset_test)
print("mAP on test dataset: {}".format(test_map[1][1]))
ag.Categorical(5e-3, 2e-3, 5e-4, 2e-4),
'batch_size':
4,
'epochs':
ag.Categorical(30, 40, 50, 60),
'nthreads_per_trial':
16,
'ngpus_per_trial':
4,
'data_shape': [640, 800],
'lr_decay_epoch':
ag.Categorical([24, 28], [35], [50, 55], [40], [45], [55],
[30, 35], [20]),
# 'warmup_epochs': 2,
'warmup_iters':
ag.Int(5, 500),
'wd':
ag.Categorical(5e-4, 2.5e-4, 1e-4),
'transfer':
transfer,
'label_smooth':
False,
'syncbn':
ag.Bool(),
'num_trials':
30,
'time_limits':
time_limits,
'dist_ip_addrs':
None
}
False, 'data_shape': (640, 800), 'nthreads_per_trial': 12, 'verbose': False, 'ngpus_per_trial': 4, 'batch_size': 4, 'hybridize': True, 'lr_decay_epoch': ag.Categorical([24, 28], [35], [50, 55], [40], [45], [55], [30, 35], [20]), 'warmup_iters': ag.Int(5, 500), 'resume': False, 'checkpoint': 'checkpoint/exp1.ag', 'visualizer': 'none', 'start_epoch': 0, 'lr_mode': 'step', 'lr_decay': 0.1, 'lr_decay_period': 0, 'warmup_lr':
parser.add_argument('--index-file-name-trainval', type=str, default='', help="name of txt file which contains images for training and validation ")
parser.add_argument('--index-file-name-test', type=str, default='', help="name of txt file which contains images for testing")
parser.add_argument('--classes', type=tuple, default=None, help="classes for custom classes")
parser.add_argument('--no-redownload', action='store_true', help="whether need to re-download dataset")
args = parser.parse_args()
logging.info('args: {}'.format(args))
dataset_train, dataset_test = get_dataset(args)
time_limits = 5*60*60 # 5 days
epochs = 1
detector = task.fit(dataset_train,
num_trials=30,
epochs=epochs,
net=ag.Categorical('darknet53', 'mobilenet1.0'),
lr=ag.Categorical(1e-2, 5e-3, 1e-3, 5e-4, 1e-4, 5e-5),
data_shape=ag.Categorical(320, 416),
ngpus_per_trial=1,
batch_size=8,
lr_decay_epoch=ag.Categorical('80,90','85,95'),
warmup_epochs=ag.Int(1, 10),
syncbn=ag.Bool(),
label_smooth=ag.Bool(),
time_limits=time_limits,
dist_ip_addrs = [])
test_map = detector.evaluate(dataset_test)
print("mAP on test dataset: {}".format(test_map[1][1]))