def test_detection_dataloader():
dataset = DummyDetectionDataset(8)
for num_workers in [0, 1, 2, 4]:
for shuffle in (True, False):
for last_batch in ('keep', 'discard', 'rollover'):
dataloader = DetectionDataLoader(dataset,
batch_size=2,
shuffle=shuffle,
last_batch=last_batch,
num_workers=num_workers)
for batch in dataloader:
mx.nd.waitall()
pass
def test_detection_dataloader():
dataset = DummyDetectionDataset(8)
for num_workers in [0, 1, 2, 4]:
for shuffle in (True, False):
for last_batch in ('keep', 'discard', 'rollover'):
dataloader = DetectionDataLoader(dataset,
batch_size=2,
shuffle=shuffle,
last_batch=last_batch,
num_workers=num_workers)
for batch in dataloader:
mx.nd.waitall()
pass
# new dataloader methods
batchify_fn = Tuple(Stack(), Pad(pad_val=-1))
dataloader = mx.gluon.data.DataLoader(dataset,
batch_size=2,
shuffle=shuffle,
last_batch=last_batch,
batchify_fn=batchify_fn,
num_workers=num_workers)
for batch in dataloader:
mx.nd.waitall()
pass
batchify_fn = Tuple(Append(), Append())
dataloader = mx.gluon.data.DataLoader(dataset,
batch_size=2,
shuffle=shuffle,
last_batch=last_batch,
batchify_fn=batchify_fn,
num_workers=num_workers)
for batch in dataloader:
mx.nd.waitall()
pass
class_names=train_dataset.classes)
plt.show()
#%% 测试转换函数，尝试将图片缩小到标准图片
train_transform = SSDDefaultTrainTransform(width, height)
train_image2, train_label2 = train_transform(train_image, train_label)
print('tensor shape:', train_image2.shape)
print('label shape:', train_label2.shape)
train_image2 = train_image2.transpose((1, 2, 0)) * nd.array(
(0.229, 0.224, 0.225)) + nd.array((0.485, 0.456, 0.406))
train_image2 = (train_image2 * 255).clip(0, 255)
ax = viz.plot_bbox(train_image2.asnumpy(),
train_label2[:, :4],
labels=train_label2[:, 4:5],
class_names=train_dataset.classes)
plt.show()
batch_size = 2 # for tutorial, we use smaller batch-size
num_workers = 0 # you can make it larger(if your CPU has more cores) to accelerate data loading
train_loader = DetectionDataLoader(train_dataset.transform(train_transform),
batch_size,
shuffle=True,
last_batch='rollover',
num_workers=num_workers)
for ib, batch in enumerate(train_loader):
if ib > 3:
break
print('data:', batch[0].shape, 'label:', batch[1].shape)
batch_size = 20
maxTrain_inOneEpoch = 1e20
epoch_num = 10
lambd = 1 / 4
Models_tmp_Dir = 'D:/Temps/Models_tmp/'
#CPU_percentage = 0.2
#使用训练元件
ctx = [mx.gpu(i) for i in range(mx.context.num_gpus())]
#ctx = [mx.cpu()]
val_transform = presets.ssd.SSDDefaultValTransform(width, height)
val_loader = DetectionDataLoader(val_dataset.transform(val_transform),
batch_size,
shuffle=False,
last_batch='keep',
num_workers=0)
#%% 网络
classes = data.VOCDetection.CLASSES
name = 'resnet50_v1'
base_size = 512
features = ['stage3_activation5', 'stage4_activation2']
filters = [512, 512, 256, 256]
sizes = [51.2, 102.4, 189.4, 276.4, 363.52, 450.6, 492]
ratios = [[1, 2, 0.5]] + [[1, 2, 0.5, 3, 1.0 / 3]] * 3 + [[1, 2, 0.5]] * 2
steps = [16, 32, 64, 128, 256, 512]
pretrained = True