deep-person-reid/torchreid/data_manager.py

from __future__ import absolute_import
from __future__ import print_function

from torch.utils.data import DataLoader

from .dataset_loader import ImageDataset, VideoDataset
from .datasets import init_imgreid_dataset, init_vidreid_dataset
from .transforms import build_transforms


class BaseDataManager(object):

    def return_dataloaders(self):
        return self.trainloader, self.testloader_dict


class ImageDataManager(BaseDataManager):
    """
    Image-ReID data manager
    """

    def __init__(self,
                 use_gpu,
                 train_names,
                 test_names,
                 root,
                 split_id=0,
                 height=256,
                 width=128,
                 train_batch_size=32,
                 test_batch_size=100,
                 workers=4,
                 cuhk03_labeled=False,
                 cuhk03_classic_split=False
                 ):
        super(ImageDataManager, self).__init__()
        pin_memory = True if use_gpu else False

        transform_train = build_transforms(height, width, is_train=True)
        transform_test = build_transforms(height, width, is_train=False)

        self.train_names = train_names
        self.test_names = test_names
        
        self.train = []
        self.num_train_pids = 0
        self.num_train_cams = 0

        print("=> Initializing TRAIN datasets")

        for name in self.train_names:
            dataset = init_imgreid_dataset(
                root=root, name=name, split_id=split_id, cuhk03_labeled=cuhk03_labeled,
                cuhk03_classic_split=cuhk03_classic_split
            )

            for img_path, pid, camid in dataset.train:
                pid += self.num_train_pids
                camid += self.num_train_cams
                self.train.append((img_path, pid, camid))

            self.num_train_pids += dataset.num_train_pids
            self.num_train_cams += dataset.num_train_cams

        self.trainloader = DataLoader(
            ImageDataset(self.train, transform=transform_train),
            batch_size=train_batch_size, shuffle=True, num_workers=workers,
            pin_memory=pin_memory, drop_last=True
        )

        print("=> Initializing TEST datasets")

        self.testloader_dict = {name: {'query': None, 'gallery': None} for name in self.test_names}
        for name in self.test_names:
            dataset = init_imgreid_dataset(
                root=root, name=name, split_id=split_id, cuhk03_labeled=cuhk03_labeled,
                cuhk03_classic_split=cuhk03_classic_split
            )

            self.testloader_dict[name]['query'] = DataLoader(
                ImageDataset(dataset.query, transform=transform_test),
                batch_size=test_batch_size, shuffle=False, num_workers=workers,
                pin_memory=pin_memory, drop_last=False
            )

            self.testloader_dict[name]['gallery'] = DataLoader(
                ImageDataset(dataset.gallery, transform=transform_test),
                batch_size=test_batch_size, shuffle=False, num_workers=workers,
                pin_memory=pin_memory, drop_last=False
            )

        print("\n")
        print("  **************** Summary ****************")
        print("  train names      : {}".format(self.train_names))
        print("  # train datasets : {}".format(len(self.train_names)))
        print("  # train ids      : {}".format(self.num_train_pids))
        print("  # train images   : {}".format(len(self.train)))
        print("  # train cameras  : {}".format(self.num_train_cams))
        print("  test names       : {}".format(self.test_names))
        print("  *****************************************")
        print("\n")


class VideoDataManager(BaseDataManager):
    """
    Video-ReID data manager
    """

    def __init__(self,
                 use_gpu,
                 train_names,
                 test_names,
                 root,
                 split_id=0,
                 height=256,
                 width=128,
                 train_batch_size=32,
                 test_batch_size=100,
                 workers=4,
                 seq_len=15,
                 sample='evenly',
                 image_training=True
                 ):
        super(VideoDataManager, self).__init__()
        pin_memory = True if use_gpu else False

        transform_train = build_transforms(height, width, is_train=True)
        transform_test = build_transforms(height, width, is_train=False)

        self.train_names = train_names
        self.test_names = test_names
        
        self.train = []
        self.num_train_pids = 0
        self.num_train_cams = 0

        print("=> Initializing TRAIN datasets")

        for name in self.train_names:
            dataset = init_vidreid_dataset(root=root, name=name, split_id=split_id)

            for img_paths, pid, camid in dataset.train:
                pid += self.num_train_pids
                camid += self.num_train_cams
                if image_training:
                    # decompose tracklets into images
                    for img_path in img_paths:
                        self.train.append((img_path, pid, camid))
                else:
                    self.train.append((img_paths, pid, camid))

            self.num_train_pids += dataset.num_train_pids
            self.num_train_cams += dataset.num_train_cams

        if image_training:
            # each batch has image data of shape (batch, channel, height, width)
            self.trainloader = DataLoader(
                ImageDataset(self.train, transform=transform_train),
                batch_size=train_batch_size, shuffle=True, num_workers=workers,
                pin_memory=pin_memory, drop_last=True
            )
        else:
            # each batch has image data of shape (batch, seq_len, channel, height, width)
            self.trainloader = DataLoader(
                VideoDataset(self.train, seq_len=seq_len, sample=sample, transform=transform_test),
                batch_size=train_batch_size, shuffle=True, num_workers=workers,
                pin_memory=pin_memory, drop_last=True
            )

        print("=> Initializing TEST datasets")

        self.testloader_dict = {name: {'query': None, 'gallery': None} for name in self.test_names}
        for name in self.test_names:
            dataset = init_vidreid_dataset(root=root, name=name, split_id=split_id)

            self.testloader_dict[name]['query'] = DataLoader(
                VideoDataset(dataset.query, seq_len=seq_len, sample=sample, transform=transform_test),
                batch_size=test_batch_size, shuffle=False, num_workers=workers,
                pin_memory=pin_memory, drop_last=False,
            )

            self.testloader_dict[name]['gallery'] = DataLoader(
                VideoDataset(dataset.gallery, seq_len=seq_len, sample=sample, transform=transform_test),
                batch_size=test_batch_size, shuffle=False, num_workers=workers,
                pin_memory=pin_memory, drop_last=False,
            )

        print("\n")
        print("  **************** Summary ****************")
        print("  train names       : {}".format(self.train_names))
        print("  # train datasets  : {}".format(len(self.train_names)))
        print("  # train ids       : {}".format(self.num_train_pids))
        if image_training:
            print("  # train images   : {}".format(len(self.train)))
        else:
            print("  # train tracklets: {}".format(len(self.train)))
        print("  # train cameras   : {}".format(self.num_train_cams))
        print("  test names        : {}".format(self.test_names))
        print("  *****************************************")
        print("\n")