[fbsync] Adding Mixup and Cutmix (#4379)

Summary:
* Add RandomMixupCutmix.

* Add test with real data.

* Use dataloader and collate in the test.

* Making RandomMixupCutmix JIT scriptable.

* Move out label_smoothing and try roll instead of flip

* Adding mixup/cutmix in references script.

* Handle one-hot encoded target in accuracy.

* Add support of devices on tests.

* Separate Mixup from Cutmix.

* Add check for floats.

* Adding device on expect value.

* Remove hardcoded weights.

* One-hot only when necessary.

* Fix linter.

* Moving mixup and cutmix to references.

* Final code clean up.

Reviewed By: datumbox

Differential Revision: D31268036

fbshipit-source-id: 6a73c079d667443da898e3b175b88978b24d52ad

references/classification/train.py

@@ -4,11 +4,13 @@
 
 import torch
 import torch.utils.data
+from torch.utils.data.dataloader import default_collate
 from torch import nn
 import torchvision
 from torchvision.transforms.functional import InterpolationMode
 
 import presets
+import transforms
 import utils
 
 try:
@@ -164,10 +166,21 @@ def main(args):
     train_dir = os.path.join(args.data_path, 'train')
     val_dir = os.path.join(args.data_path, 'val')
     dataset, dataset_test, train_sampler, test_sampler = load_data(train_dir, val_dir, args)
+
+    collate_fn = None
+    num_classes = len(dataset.classes)
+    mixup_transforms = []
+    if args.mixup_alpha > 0.0:
+        mixup_transforms.append(transforms.RandomMixup(num_classes, p=1.0, alpha=args.mixup_alpha))
+    if args.cutmix_alpha > 0.0:
+        mixup_transforms.append(transforms.RandomCutmix(num_classes, p=1.0, alpha=args.cutmix_alpha))
+    if mixup_transforms:
+        mixupcutmix = torchvision.transforms.RandomChoice(mixup_transforms)
+        collate_fn = lambda batch: mixupcutmix(*default_collate(batch))  # noqa: E731
     data_loader = torch.utils.data.DataLoader(
         dataset, batch_size=args.batch_size,
-        sampler=train_sampler, num_workers=args.workers, pin_memory=True)
-
+        sampler=train_sampler, num_workers=args.workers, pin_memory=True,
+        collate_fn=collate_fn)
     data_loader_test = torch.utils.data.DataLoader(
         dataset_test, batch_size=args.batch_size,
         sampler=test_sampler, num_workers=args.workers, pin_memory=True)
@@ -272,6 +285,8 @@ def get_args_parser(add_help=True):
     parser.add_argument('--label-smoothing', default=0.0, type=float,
                         help='label smoothing (default: 0.0)',
                         dest='label_smoothing')
+    parser.add_argument('--mixup-alpha', default=0.0, type=float, help='mixup alpha (default: 0.0)')
+    parser.add_argument('--cutmix-alpha', default=0.0, type=float, help='cutmix alpha (default: 0.0)')
     parser.add_argument('--lr-step-size', default=30, type=int, help='decrease lr every step-size epochs')
     parser.add_argument('--lr-gamma', default=0.1, type=float, help='decrease lr by a factor of lr-gamma')
     parser.add_argument('--print-freq', default=10, type=int, help='print frequency')

references/classification/transforms.py

@@ -0,0 +1,175 @@
+import math
+import torch
+
+from typing import Tuple
+from torch import Tensor
+from torchvision.transforms import functional as F
+
+
+class RandomMixup(torch.nn.Module):
+    """Randomly apply Mixup to the provided batch and targets.
+    The class implements the data augmentations as described in the paper
+    `"mixup: Beyond Empirical Risk Minimization" <https://arxiv.org/abs/1710.09412>`_.
+
+    Args:
+        num_classes (int): number of classes used for one-hot encoding.
+        p (float): probability of the batch being transformed. Default value is 0.5.
+        alpha (float): hyperparameter of the Beta distribution used for mixup.
+            Default value is 1.0.
+        inplace (bool): boolean to make this transform inplace. Default set to False.
+    """
+
+    def __init__(self, num_classes: int,
+                 p: float = 0.5, alpha: float = 1.0,
+                 inplace: bool = False) -> None:
+        super().__init__()
+        assert num_classes > 0, "Please provide a valid positive value for the num_classes."
+        assert alpha > 0, "Alpha param can't be zero."
+
+        self.num_classes = num_classes
+        self.p = p
+        self.alpha = alpha
+        self.inplace = inplace
+
+    def forward(self, batch: Tensor, target: Tensor) -> Tuple[Tensor, Tensor]:
+        """
+        Args:
+            batch (Tensor): Float tensor of size (B, C, H, W)
+            target (Tensor): Integer tensor of size (B, )
+
+        Returns:
+            Tensor: Randomly transformed batch.
+        """
+        if batch.ndim != 4:
+            raise ValueError("Batch ndim should be 4. Got {}".format(batch.ndim))
+        elif target.ndim != 1:
+            raise ValueError("Target ndim should be 1. Got {}".format(target.ndim))
+        elif not batch.is_floating_point():
+            raise TypeError('Batch dtype should be a float tensor. Got {}.'.format(batch.dtype))
+        elif target.dtype != torch.int64:
+            raise TypeError("Target dtype should be torch.int64. Got {}".format(target.dtype))
+
+        if not self.inplace:
+            batch = batch.clone()
+            target = target.clone()
+
+        if target.ndim == 1:
+            target = torch.nn.functional.one_hot(target, num_classes=self.num_classes).to(dtype=batch.dtype)
+
+        if torch.rand(1).item() >= self.p:
+            return batch, target
+
+        # It's faster to roll the batch by one instead of shuffling it to create image pairs
+        batch_rolled = batch.roll(1, 0)
+        target_rolled = target.roll(1, 0)
+
+        # Implemented as on mixup paper, page 3.
+        lambda_param = float(torch._sample_dirichlet(torch.tensor([self.alpha, self.alpha]))[0])
+        batch_rolled.mul_(1.0 - lambda_param)
+        batch.mul_(lambda_param).add_(batch_rolled)
+
+        target_rolled.mul_(1.0 - lambda_param)
+        target.mul_(lambda_param).add_(target_rolled)
+
+        return batch, target
+
+    def __repr__(self) -> str:
+        s = self.__class__.__name__ + '('
+        s += 'num_classes={num_classes}'
+        s += ', p={p}'
+        s += ', alpha={alpha}'
+        s += ', inplace={inplace}'
+        s += ')'
+        return s.format(**self.__dict__)
+
+
+class RandomCutmix(torch.nn.Module):
+    """Randomly apply Cutmix to the provided batch and targets.
+    The class implements the data augmentations as described in the paper
+    `"CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features"
+    <https://arxiv.org/abs/1905.04899>`_.
+
+    Args:
+        num_classes (int): number of classes used for one-hot encoding.
+        p (float): probability of the batch being transformed. Default value is 0.5.
+        alpha (float): hyperparameter of the Beta distribution used for cutmix.
+            Default value is 1.0.
+        inplace (bool): boolean to make this transform inplace. Default set to False.
+    """
+
+    def __init__(self, num_classes: int,
+                 p: float = 0.5, alpha: float = 1.0,
+                 inplace: bool = False) -> None:
+        super().__init__()
+        assert num_classes > 0, "Please provide a valid positive value for the num_classes."
+        assert alpha > 0, "Alpha param can't be zero."
+
+        self.num_classes = num_classes
+        self.p = p
+        self.alpha = alpha
+        self.inplace = inplace
+
+    def forward(self, batch: Tensor, target: Tensor) -> Tuple[Tensor, Tensor]:
+        """
+        Args:
+            batch (Tensor): Float tensor of size (B, C, H, W)
+            target (Tensor): Integer tensor of size (B, )
+
+        Returns:
+            Tensor: Randomly transformed batch.
+        """
+        if batch.ndim != 4:
+            raise ValueError("Batch ndim should be 4. Got {}".format(batch.ndim))
+        elif target.ndim != 1:
+            raise ValueError("Target ndim should be 1. Got {}".format(target.ndim))
+        elif not batch.is_floating_point():
+            raise TypeError('Batch dtype should be a float tensor. Got {}.'.format(batch.dtype))
+        elif target.dtype != torch.int64:
+            raise TypeError("Target dtype should be torch.int64. Got {}".format(target.dtype))
+
+        if not self.inplace:
+            batch = batch.clone()
+            target = target.clone()
+
+        if target.ndim == 1:
+            target = torch.nn.functional.one_hot(target, num_classes=self.num_classes).to(dtype=batch.dtype)
+
+        if torch.rand(1).item() >= self.p:
+            return batch, target
+
+        # It's faster to roll the batch by one instead of shuffling it to create image pairs
+        batch_rolled = batch.roll(1, 0)
+        target_rolled = target.roll(1, 0)
+
+        # Implemented as on cutmix paper, page 12 (with minor corrections on typos).
+        lambda_param = float(torch._sample_dirichlet(torch.tensor([self.alpha, self.alpha]))[0])
+        W, H = F.get_image_size(batch)
+
+        r_x = torch.randint(W, (1,))
+        r_y = torch.randint(H, (1,))
+
+        r = 0.5 * math.sqrt(1.0 - lambda_param)
+        r_w_half = int(r * W)
+        r_h_half = int(r * H)
+
+        x1 = int(torch.clamp(r_x - r_w_half, min=0))
+        y1 = int(torch.clamp(r_y - r_h_half, min=0))
+        x2 = int(torch.clamp(r_x + r_w_half, max=W))
+        y2 = int(torch.clamp(r_y + r_h_half, max=H))
+
+        batch[:, :, y1:y2, x1:x2] = batch_rolled[:, :, y1:y2, x1:x2]
+        lambda_param = float(1.0 - (x2 - x1) * (y2 - y1) / (W * H))
+
+        target_rolled.mul_(1.0 - lambda_param)
+        target.mul_(lambda_param).add_(target_rolled)
+
+        return batch, target
+
+    def __repr__(self) -> str:
+        s = self.__class__.__name__ + '('
+        s += 'num_classes={num_classes}'
+        s += ', p={p}'
+        s += ', alpha={alpha}'
+        s += ', inplace={inplace}'
+        s += ')'
+        return s.format(**self.__dict__)

references/classification/utils.py

@@ -189,6 +189,8 @@ def accuracy(output, target, topk=(1,)):
     with torch.no_grad():
         maxk = max(topk)
         batch_size = target.size(0)
+        if target.ndim == 2:
+            target = target.max(dim=1)[1]
 
         _, pred = output.topk(maxk, 1, True, True)
         pred = pred.t()

test/test_transforms.py

@@ -1311,7 +1311,8 @@ def test_random_choice():
             transforms.Resize(15),
             transforms.Resize(20),
             transforms.CenterCrop(10)
-        ]
+        ],
+        [1 / 3, 1 / 3, 1 / 3]
     )
     img = transforms.ToPILImage()(torch.rand(3, 25, 25))
     num_samples = 250

torchvision/transforms/transforms.py

@@ -515,9 +515,20 @@ def __call__(self, img):
 class RandomChoice(RandomTransforms):
     """Apply single transformation randomly picked from a list. This transform does not support torchscript.
     """
-    def __call__(self, img):
-        t = random.choice(self.transforms)
-        return t(img)
+    def __init__(self, transforms, p=None):
+        super().__init__(transforms)
+        if p is not None and not isinstance(p, Sequence):
+            raise TypeError("Argument transforms should be a sequence")
+        self.p = p
+
+    def __call__(self, *args):
+        t = random.choices(self.transforms, weights=self.p)[0]
+        return t(*args)
+
+    def __repr__(self):
+        format_string = super().__repr__()
+        format_string += '(p={0})'.format(self.p)
+        return format_string
 
 
 class RandomCrop(torch.nn.Module):