def prepare_data_loader(datapth, annpath, inputsize, imgs_per_gpu, gpu_count, scales, cropsize, anns_ignore=255, mode='train', distributed=False):
if mode == 'train':
transforms = TransformationTrain(scales, cropsize)
batchsize = imgs_per_gpu * gpu_count
shuffle = True
drop_last = True
elif mode == 'val':
transforms = TransformationVal()
batchsize = imgs_per_gpu * gpu_count
shuffle = False
drop_last = False
elif mode == 'test':
transforms = TransformationVal()
batchsize = 1
shuffle = False
drop_last = False
else: assert mode is None, "mode should be defined"
ds = Carla(datapth, annpath, inputsize, tar_ignore=anns_ignore, transforms=transforms, mode=mode)
dl = DataLoader(
ds,
batch_size=batchsize,
shuffle=shuffle,
drop_last=drop_last,
num_workers=4,
pin_memory=True,
)
return dl
def get_data_loader(datapth,
annpath,
ims_per_gpu,
scales,
cropsize,
mode='train'): #, distributed=True):
if mode == 'train':
trans_func = TransformationTrain(scales, cropsize)
batchsize = ims_per_gpu
#batchsize = 8
shuffle = True
drop_last = True
elif mode == 'val':
trans_func = TransformationVal()
batchsize = ims_per_gpu
shuffle = False
drop_last = False
dss = CityScapes(datapth, annpath, trans_func=trans_func, mode=mode)
count = 0
dl = DataLoader(dss,
batch_size=batchsize,
shuffle=shuffle,
drop_last=drop_last,
num_workers=4,
pin_memory=True)
return dl
def get_data_loader(datapth,
annpath,
ims_per_gpu,
scales,
cropsize,
max_iter=None,
mode='train',
distributed=True):
if mode == 'train':
trans_func = TransformationTrain(scales, cropsize)
batchsize = ims_per_gpu
shuffle = True
drop_last = True
elif mode == 'val':
trans_func = TransformationVal(cropsize)
batchsize = ims_per_gpu
shuffle = False
drop_last = False
ds = Coco(datapth, annpath, trans_func=trans_func, mode=mode)
if distributed:
assert dist.is_available(), "dist should be initialzed"
if mode == 'train':
assert not max_iter is None
n_train_imgs = ims_per_gpu * dist.get_world_size() * max_iter
sampler = RepeatedDistSampler(ds, n_train_imgs, shuffle=shuffle)
else:
sampler = torch.utils.data.distributed.DistributedSampler(
ds, shuffle=shuffle)
batchsampler = torch.utils.data.sampler.BatchSampler(
sampler, batchsize, drop_last=drop_last)
dl = DataLoader(
ds,
batch_sampler=batchsampler,
num_workers=4,
pin_memory=True,
)
else:
dl = DataLoader(
ds,
batch_size=batchsize,
shuffle=shuffle,
drop_last=drop_last,
num_workers=4,
pin_memory=True,
)
return dl