Add the S3D architecture to TorchVision (#6412)

* S3D initial commit

* add model builder code and docstrings

* change classifier submodule, populate weights enum

* fix change of block args from List[List[int]] to ints

* add VideoClassification to transforms

* edit weights url for testing, add s3d to models.video init

* norm_layer changes

* norm_layer and args fix

* Overwrite default dropout

* Remove docs from internal submodules.

* Fix tests

* Adding documentation.

* Link doc from main models.rst

* Fix min_temporal_size

* Adding crop/resize parameters in references script

* Release weights.

* Refactor dropout.

* Adding the weights table in the doc

Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>
Co-authored-by: Vasilis Vryniotis <vvryniotis@fb.com>

docs/source/models.rst

@@ -516,6 +516,7 @@ pre-trained weights:
 
    models/video_mvit
    models/video_resnet
+   models/video_s3d
 
 |
 

docs/source/models/video_s3d.rst

@@ -0,0 +1,25 @@
+Video S3D
+=========
+
+.. currentmodule:: torchvision.models.video
+
+The S3D model is based on the
+`Rethinking Spatiotemporal Feature Learning: Speed-Accuracy Trade-offs in Video Classification
+<https://arxiv.org/abs/1712.04851>`__ paper.
+
+
+Model builders
+--------------
+
+The following model builders can be used to instantiate an S3D model, with or
+without pre-trained weights. All the model builders internally rely on the
+``torchvision.models.video.S3D`` base class. Please refer to the `source
+code
+<https://github.com/pytorch/vision/blob/main/torchvision/models/video/s3d.py>`_ for
+more details about this class.
+
+.. autosummary::
+    :toctree: generated/
+    :template: function.rst
+
+    s3d

test/test_models.py

@@ -312,6 +312,9 @@ def _check_input_backprop(model, inputs):
     "mvit_v2_s": {
         "input_shape": (1, 3, 16, 224, 224),
     },
+    "s3d": {
+        "input_shape": (1, 3, 16, 224, 224),
+    },
 }
 # speeding up slow models:
 slow_models = [

torchvision/models/video/__init__.py

@@ -1,2 +1,3 @@
 from .mvit import *
 from .resnet import *
+from .s3d import *

torchvision/models/video/s3d.py

@@ -0,0 +1,216 @@
+from functools import partial
+from typing import Any, Callable, Optional
+
+import torch
+from torch import nn
+from torchvision.ops.misc import Conv3dNormActivation
+
+from ...transforms._presets import VideoClassification
+from ...utils import _log_api_usage_once
+from .._api import register_model, Weights, WeightsEnum
+from .._meta import _KINETICS400_CATEGORIES
+from .._utils import _ovewrite_named_param
+
+
+__all__ = [
+    "S3D",
+    "S3D_Weights",
+    "s3d",
+]
+
+
+class TemporalSeparableConv(nn.Sequential):
+    def __init__(
+        self,
+        in_planes: int,
+        out_planes: int,
+        kernel_size: int,
+        stride: int,
+        padding: int,
+        norm_layer: Callable[..., nn.Module],
+    ):
+        super().__init__(
+            Conv3dNormActivation(
+                in_planes,
+                out_planes,
+                kernel_size=(1, kernel_size, kernel_size),
+                stride=(1, stride, stride),
+                padding=(0, padding, padding),
+                bias=False,
+                norm_layer=norm_layer,
+            ),
+            Conv3dNormActivation(
+                out_planes,
+                out_planes,
+                kernel_size=(kernel_size, 1, 1),
+                stride=(stride, 1, 1),
+                padding=(padding, 0, 0),
+                bias=False,
+                norm_layer=norm_layer,
+            ),
+        )
+
+
+class SepInceptionBlock3D(nn.Module):
+    def __init__(
+        self,
+        in_planes: int,
+        b0_out: int,
+        b1_mid: int,
+        b1_out: int,
+        b2_mid: int,
+        b2_out: int,
+        b3_out: int,
+        norm_layer: Callable[..., nn.Module],
+    ):
+        super().__init__()
+
+        self.branch0 = Conv3dNormActivation(in_planes, b0_out, kernel_size=1, stride=1, norm_layer=norm_layer)
+        self.branch1 = nn.Sequential(
+            Conv3dNormActivation(in_planes, b1_mid, kernel_size=1, stride=1, norm_layer=norm_layer),
+            TemporalSeparableConv(b1_mid, b1_out, kernel_size=3, stride=1, padding=1, norm_layer=norm_layer),
+        )
+        self.branch2 = nn.Sequential(
+            Conv3dNormActivation(in_planes, b2_mid, kernel_size=1, stride=1, norm_layer=norm_layer),
+            TemporalSeparableConv(b2_mid, b2_out, kernel_size=3, stride=1, padding=1, norm_layer=norm_layer),
+        )
+        self.branch3 = nn.Sequential(
+            nn.MaxPool3d(kernel_size=(3, 3, 3), stride=1, padding=1),
+            Conv3dNormActivation(in_planes, b3_out, kernel_size=1, stride=1, norm_layer=norm_layer),
+        )
+
+    def forward(self, x):
+        x0 = self.branch0(x)
+        x1 = self.branch1(x)
+        x2 = self.branch2(x)
+        x3 = self.branch3(x)
+        out = torch.cat((x0, x1, x2, x3), 1)
+
+        return out
+
+
+class S3D(nn.Module):
+    """S3D main class.
+
+    Args:
+        num_class (int): number of classes for the classification task.
+        dropout (float): dropout probability.
+        norm_layer (Optional[Callable]): Module specifying the normalization layer to use.
+
+    Inputs:
+        x (Tensor): batch of videos with dimensions (batch, channel, time, height, width)
+    """
+
+    def __init__(
+        self,
+        num_classes: int = 400,
+        dropout: float = 0.0,
+        norm_layer: Optional[Callable[..., torch.nn.Module]] = None,
+    ) -> None:
+        super().__init__()
+        _log_api_usage_once(self)
+
+        if norm_layer is None:
+            norm_layer = partial(nn.BatchNorm3d, eps=0.001, momentum=0.001)
+
+        self.features = nn.Sequential(
+            TemporalSeparableConv(3, 64, 7, 2, 3, norm_layer),
+            nn.MaxPool3d(kernel_size=(1, 3, 3), stride=(1, 2, 2), padding=(0, 1, 1)),
+            Conv3dNormActivation(
+                64,
+                64,
+                kernel_size=1,
+                stride=1,
+                norm_layer=norm_layer,
+            ),
+            TemporalSeparableConv(64, 192, 3, 1, 1, norm_layer),
+            nn.MaxPool3d(kernel_size=(1, 3, 3), stride=(1, 2, 2), padding=(0, 1, 1)),
+            SepInceptionBlock3D(192, 64, 96, 128, 16, 32, 32, norm_layer),
+            SepInceptionBlock3D(256, 128, 128, 192, 32, 96, 64, norm_layer),
+            nn.MaxPool3d(kernel_size=(3, 3, 3), stride=(2, 2, 2), padding=(1, 1, 1)),
+            SepInceptionBlock3D(480, 192, 96, 208, 16, 48, 64, norm_layer),
+            SepInceptionBlock3D(512, 160, 112, 224, 24, 64, 64, norm_layer),
+            SepInceptionBlock3D(512, 128, 128, 256, 24, 64, 64, norm_layer),
+            SepInceptionBlock3D(512, 112, 144, 288, 32, 64, 64, norm_layer),
+            SepInceptionBlock3D(528, 256, 160, 320, 32, 128, 128, norm_layer),
+            nn.MaxPool3d(kernel_size=(2, 2, 2), stride=(2, 2, 2), padding=(0, 0, 0)),
+            SepInceptionBlock3D(832, 256, 160, 320, 32, 128, 128, norm_layer),
+            SepInceptionBlock3D(832, 384, 192, 384, 48, 128, 128, norm_layer),
+        )
+        self.avgpool = nn.AvgPool3d(kernel_size=(2, 7, 7), stride=1)
+        self.classifier = nn.Sequential(
+            nn.Dropout(p=dropout),
+            nn.Conv3d(1024, num_classes, kernel_size=1, stride=1, bias=True),
+        )
+
+    def forward(self, x):
+        x = self.features(x)
+        x = self.avgpool(x)
+        x = self.classifier(x)
+        x = torch.mean(x, dim=(2, 3, 4))
+        return x
+
+
+class S3D_Weights(WeightsEnum):
+    KINETICS400_V1 = Weights(
+        url="https://download.pytorch.org/models/s3d-1bd8ae63.pth",
+        transforms=partial(
+            VideoClassification,
+            crop_size=(224, 224),
+            resize_size=(256, 256),
+            mean=(0.5, 0.5, 0.5),
+            std=(0.5, 0.5, 0.5),
+        ),
+        meta={
+            "min_size": (224, 224),
+            "min_temporal_size": 14,
+            "categories": _KINETICS400_CATEGORIES,
+            "recipe": "https://github.com/pytorch/vision/pull/6412#issuecomment-1219687434",
+            "_docs": (
+                "The weights are ported from a community repository. The accuracies are estimated on clip-level "
+                "with parameters `frame_rate=15`, `clips_per_video=1`, and `clip_len=128`."
+            ),
+            "num_params": 8320048,
+            "_metrics": {
+                "Kinetics-400": {
+                    "acc@1": 67.315,
+                    "acc@5": 87.593,
+                }
+            },
+        },
+    )
+    DEFAULT = KINETICS400_V1
+
+
+@register_model()
+def s3d(*, weights: Optional[S3D_Weights] = None, progress: bool = True, **kwargs: Any) -> S3D:
+    """Construct Separable 3D CNN model.
+
+    Reference: `Rethinking Spatiotemporal Feature Learning <https://arxiv.org/abs/1712.04851>`__.
+
+    Args:
+        weights (:class:`~torchvision.models.video.S3D_Weights`, optional): The
+            pretrained weights to use. See
+            :class:`~torchvision.models.video.S3D_Weights`
+            below for more details, and possible values. By default, no
+            pre-trained weights are used.
+        progress (bool): If True, displays a progress bar of the download to stderr. Default is True.
+        **kwargs: parameters passed to the ``torchvision.models.video.S3D`` base class.
+            Please refer to the `source code
+            <https://github.com/pytorch/vision/blob/main/torchvision/models/video/s3d.py>`_
+            for more details about this class.
+
+    .. autoclass:: torchvision.models.video.S3D_Weights
+        :members:
+    """
+    weights = S3D_Weights.verify(weights)
+
+    if weights is not None:
+        _ovewrite_named_param(kwargs, "num_classes", len(weights.meta["categories"]))
+
+    model = S3D(**kwargs)
+
+    if weights is not None:
+        model.load_state_dict(weights.get_state_dict(progress=progress))
+
+    return model