ORBIT-Dataset/data/queues.py

80 строки
3.7 KiB
Python

# Copyright (c) Microsoft Corporation.
# Licensed under the MIT license.
import torch
from data.samplers import TaskSampler
from data.datasets import UserEpisodicORBITDataset, ObjectEpisodicORBITDataset
class DatasetQueue:
"""
Class for a queue of tasks sampled from UserEpisodicORIBTDataset/ObjectEpisodicORBITDataset.
"""
def __init__(self, num_tasks: int, shuffle: bool, num_workers: int) -> None:
"""
Creates instance of DatasetQueue.
:param num_tasks: (int) Number of tasks per user/object to add to the queue.
:param shuffle: (bool) If True, shuffle tasks, else do not shuffle.
:param num_workers: (int) Number of workers to use.
:return: Nothing.
"""
self.num_tasks = num_tasks
self.shuffle = shuffle
self.num_workers = num_workers
self.num_users = None
self.collate_fn = self.unpack
def unpack(self, batch):
#assumes batch_size = 1
assert len(batch) == 1, "DataLoader needs a batch size of 1!"
unpacked_batch = {}
for k,v in batch[0].items():
unpacked_batch[k] = v
return unpacked_batch
def get_num_users(self):
return self.num_users
def get_cluster_classes(self):
return self.dataset.cluster_classes
class UserEpisodicDatasetQueue(DatasetQueue):
def __init__(self, root, way_method, object_cap, shot_method, shots, video_types, subsample_factor, clip_methods, clip_length, frame_size, frame_norm_method, annotations_to_load, filter_by_annotations, num_tasks, test_mode, with_cluster_labels, with_caps, shuffle, num_workers=None, logfile=None):
num_workers = num_workers if num_workers else 4 if test_mode else 8
DatasetQueue.__init__(self, num_tasks, shuffle, num_workers)
self.dataset = UserEpisodicORBITDataset(root, way_method, object_cap, shot_method, shots, video_types, subsample_factor, clip_methods, clip_length, frame_size, frame_norm_method, annotations_to_load, filter_by_annotations, test_mode, with_cluster_labels, with_caps, logfile)
self.num_users = self.dataset.num_users
def get_tasks(self):
return torch.utils.data.DataLoader(
dataset=self.dataset,
pin_memory=False,
num_workers=self.num_workers,
sampler=TaskSampler(self.num_tasks, self.num_users, self.shuffle),
collate_fn=self.collate_fn
)
def __len__(self):
return self.dataset.num_users
class ObjectEpisodicDatasetQueue(DatasetQueue):
def __init__(self, root, way_method, object_cap, shot_method, shots, video_types, subsample_factor, clip_methods, clip_length, frame_size, frame_norm_method, annotations_to_load, filter_by_annotations, num_tasks, test_mode, with_cluster_labels, with_caps, shuffle, num_workers=None, logfile=None):
num_workers = num_workers if num_workers else 4 if test_mode else 8
DatasetQueue.__init__(self, num_tasks, shuffle, num_workers)
self.dataset = ObjectEpisodicORBITDataset(root, way_method, object_cap, shot_method, shots, video_types, subsample_factor, clip_methods, clip_length, frame_size, frame_norm_method, annotations_to_load, filter_by_annotations, test_mode, with_cluster_labels, with_caps, logfile)
self.num_users = self.dataset.num_users
self.num_objects = self.dataset.num_objects
def get_tasks(self):
return torch.utils.data.DataLoader(
dataset=self.dataset,
pin_memory=False,
num_workers=self.num_workers,
sampler=TaskSampler(self.num_tasks, self.num_objects, self.shuffle),
collate_fn=self.collate_fn
)
def __len__(self):
return self.dataset.num_objects