Make feature extractor pluggable

Summary: Make ResNetFeatureExtractor be an implementation of FeatureExtractorBase.

Reviewed By: davnov134

Differential Revision: D35433098

fbshipit-source-id: 0664a9166a88e150231cfe2eceba017ae55aed3a

projects/implicitron_trainer/README.md

@@ -224,7 +224,8 @@ generic_model_args: GenericModel
         └-- hypernet_args: SRNRaymarchHyperNet
         └-- pixel_generator_args: SRNPixelGenerator
     ╘== IdrFeatureField
-└-- image_feature_extractor_args: ResNetFeatureExtractor
+└-- image_feature_extractor_*_args: FeatureExtractorBase
+    ╘== ResNetFeatureExtractor
 └-- view_sampler_args: ViewSampler
 └-- feature_aggregator_*_args: FeatureAggregatorBase
     ╘== IdentityFeatureAggregator

projects/implicitron_trainer/configs/repro_base.yaml

@@ -64,7 +64,7 @@ generic_model_args:
   view_pooler_args:
     view_sampler_args:
       masked_sampling: false
-  image_feature_extractor_args:
+  image_feature_extractor_ResNetFeatureExtractor_args:
     stages:
     - 1
     - 2

projects/implicitron_trainer/configs/repro_feat_extractor_normed.yaml

@@ -1,6 +1,6 @@
 generic_model_args:
-  image_feature_extractor_enabled: true
+  image_feature_extractor_class_type: ResNetFeatureExtractor
-  image_feature_extractor_args:
+  image_feature_extractor_ResNetFeatureExtractor_args:
     add_images: true
     add_masks: true
     first_max_pool: true

projects/implicitron_trainer/configs/repro_feat_extractor_transformer.yaml

@@ -1,6 +1,6 @@
 generic_model_args:
-  image_feature_extractor_enabled: true
+  image_feature_extractor_class_type: ResNetFeatureExtractor
-  image_feature_extractor_args:
+  image_feature_extractor_ResNetFeatureExtractor_args:
     add_images: true
     add_masks: true
     first_max_pool: false

projects/implicitron_trainer/configs/repro_feat_extractor_unnormed.yaml

@@ -1,6 +1,6 @@
 generic_model_args:
-  image_feature_extractor_enabled: true
+  image_feature_extractor_class_type: ResNetFeatureExtractor
-  image_feature_extractor_args:
+  image_feature_extractor_ResNetFeatureExtractor_args:
     stages:
     - 1
     - 2

pytorch3d/implicitron/models/feature_extractor/__init__.py

@@ -0,0 +1,7 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+from .feature_extractor import FeatureExtractorBase

pytorch3d/implicitron/models/feature_extractor/feature_extractor.py

@@ -0,0 +1,44 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+from typing import Any, Dict, Optional
+
+import torch
+from pytorch3d.implicitron.tools.config import ReplaceableBase
+
+
+class FeatureExtractorBase(ReplaceableBase, torch.nn.Module):
+    """
+    Base class for an extractor of a set of features from images.
+    """
+
+    def __init__(self):
+        super().__init__()
+
+    def get_feat_dims(self) -> int:
+        """
+        Returns:
+            total number of feature dimensions of the output.
+            (i.e. sum_i(dim_i))
+        """
+        raise NotImplementedError
+
+    def forward(
+        self,
+        imgs: Optional[torch.Tensor],
+        masks: Optional[torch.Tensor] = None,
+        **kwargs
+    ) -> Dict[Any, torch.Tensor]:
+        """
+        Args:
+            imgs: A batch of input images of shape `(B, 3, H, W)`.
+            masks: A batch of input masks of shape `(B, 3, H, W)`.
+
+        Returns:
+            out_feats: A dict `{f_i: t_i}` keyed by predicted feature names `f_i`
+                and their corresponding tensors `t_i` of shape `(B, dim_i, H_i, W_i)`.
+        """
+        raise NotImplementedError

pytorch3d/implicitron/models/feature_extractor/resnet_feature_extractor.py

@@ -4,7 +4,6 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
 
-import copy
 import logging
 import math
 from typing import Any, Dict, Optional, Tuple
@@ -12,7 +11,9 @@ from typing import Any, Dict, Optional, Tuple
 import torch
 import torch.nn.functional as Fu
 import torchvision
-from pytorch3d.implicitron.tools.config import Configurable
+from pytorch3d.implicitron.tools.config import registry
+
+from . import FeatureExtractorBase
 
 
 logger = logging.getLogger(__name__)
@@ -32,7 +33,8 @@ _RESNET_MEAN = [0.485, 0.456, 0.406]
 _RESNET_STD = [0.229, 0.224, 0.225]
 
 
-class ResNetFeatureExtractor(Configurable, torch.nn.Module):
+@registry.register
+class ResNetFeatureExtractor(FeatureExtractorBase):
     """
     Implements an image feature extractor. Depending on the settings allows
     to extract:
@@ -142,14 +144,14 @@ class ResNetFeatureExtractor(Configurable, torch.nn.Module):
         return (img - self._resnet_mean) / self._resnet_std
 
     def get_feat_dims(self) -> int:
-        return (
+        # pyre-fixme[29]
-            sum(self._feat_dim.values())  # pyre-fixme[29]
+        return sum(self._feat_dim.values())
-            if len(self._feat_dim) > 0  # pyre-fixme[6]
-            else 0
-        )
 
     def forward(
-        self, imgs: torch.Tensor, masks: Optional[torch.Tensor] = None
+        self,
+        imgs: Optional[torch.Tensor],
+        masks: Optional[torch.Tensor] = None,
+        **kwargs,
     ) -> Dict[Any, torch.Tensor]:
         """
         Args:
@@ -164,7 +166,7 @@ class ResNetFeatureExtractor(Configurable, torch.nn.Module):
         out_feats = {}
 
         imgs_input = imgs
-        if self.image_rescale != 1.0:
+        if self.image_rescale != 1.0 and imgs_input is not None:
             imgs_resized = Fu.interpolate(
                 imgs_input,
                 # pyre-fixme[6]: For 2nd param expected `Optional[List[float]]` but
@@ -175,12 +177,13 @@ class ResNetFeatureExtractor(Configurable, torch.nn.Module):
         else:
             imgs_resized = imgs_input
 
-        if self.normalize_image:
-            imgs_normed = self._resnet_normalize_image(imgs_resized)
-        else:
-            imgs_normed = imgs_resized
-
         if len(self.stages) > 0:
+            assert imgs_resized is not None
+
+            if self.normalize_image:
+                imgs_normed = self._resnet_normalize_image(imgs_resized)
+            else:
+                imgs_normed = imgs_resized
             # pyre-fixme[29]: `Union[torch.Tensor, torch.nn.modules.module.Module]`
             #  is not a function.
             feats = self.stem(imgs_normed)
@@ -207,7 +210,7 @@ class ResNetFeatureExtractor(Configurable, torch.nn.Module):
             out_feats[MASK_FEATURE_NAME] = masks
 
         if self.add_images:
-            assert imgs_input is not None
+            assert imgs_resized is not None
             out_feats[IMAGE_FEATURE_NAME] = imgs_resized
 
         if self.feature_rescale != 1.0:

pytorch3d/implicitron/models/generic_model.py

@@ -5,6 +5,9 @@
 # LICENSE file in the root directory of this source tree.
 
 
+# Note: The #noqa comments below are for unused imports of pluggable implementations
+# which are part of implicitron. They ensure that the registry is prepopulated.
+
 import logging
 import math
 import warnings
@@ -27,6 +30,8 @@ from visdom import Visdom
 
 from .autodecoder import Autodecoder
 from .base_model import ImplicitronModelBase, ImplicitronRender
+from .feature_extractor import FeatureExtractorBase
+from .feature_extractor.resnet_feature_extractor import ResNetFeatureExtractor  # noqa
 from .implicit_function.base import ImplicitFunctionBase
 from .implicit_function.idr_feature_field import IdrFeatureField  # noqa
 from .implicit_function.neural_radiance_field import (  # noqa
@@ -49,7 +54,6 @@ from .renderer.lstm_renderer import LSTMRenderer  # noqa
 from .renderer.multipass_ea import MultiPassEmissionAbsorptionRenderer  # noqa
 from .renderer.ray_sampler import RaySamplerBase
 from .renderer.sdf_renderer import SignedDistanceFunctionRenderer  # noqa
-from .resnet_feature_extractor import ResNetFeatureExtractor
 from .view_pooler.view_pooler import ViewPooler
 
 
@@ -139,9 +143,6 @@ class GenericModel(ImplicitronModelBase, torch.nn.Module):  # pyre-ignore: 13
         output_rasterized_mc: If True, visualize the Monte-Carlo pixel renders by
             splatting onto an image grid. Default: False.
         bg_color: RGB values for the background color. Default (0.0, 0.0, 0.0)
-        view_pool: If True, features are sampled from the source image(s)
-            at the projected 2d locations of the sampled 3d ray points from the target
-            view(s), i.e. this activates step (3) above.
         num_passes: The specified implicit_function is initialized num_passes
             times and run sequentially.
         chunk_size_grid: The total number of points which can be rendered
@@ -169,10 +170,13 @@ class GenericModel(ImplicitronModelBase, torch.nn.Module):  # pyre-ignore: 13
             registry.
         renderer: A renderer class which inherits from BaseRenderer. This is used to
             generate the images from the target view(s).
-        image_feature_extractor_enabled: If `True`, constructs and enables
+        image_feature_extractor_class_type: If a str, constructs and enables
-            the `image_feature_extractor` object.
+            the `image_feature_extractor` object of this type. Or None if not needed.
         image_feature_extractor: A module for extrating features from an input image.
         view_pooler_enabled: If `True`, constructs and enables the `view_pooler` object.
+            This means features are sampled from the source image(s)
+            at the projected 2d locations of the sampled 3d ray points from the target
+            view(s), i.e. this activates step (3) above.
         view_pooler: An instance of ViewPooler which is used for sampling of
             image-based features at the 2D projections of a set
             of 3D points and aggregating the sampled features.
@@ -215,8 +219,9 @@ class GenericModel(ImplicitronModelBase, torch.nn.Module):  # pyre-ignore: 13
     renderer: BaseRenderer
 
     # ---- image feature extractor settings
-    image_feature_extractor_enabled: bool = False
+    # (This is only created if view_pooler is enabled)
-    image_feature_extractor: Optional[ResNetFeatureExtractor]
+    image_feature_extractor: Optional[FeatureExtractorBase]
+    image_feature_extractor_class_type: Optional[str] = "ResNetFeatureExtractor"
     # ---- view pooler settings
     view_pooler_enabled: bool = False
     view_pooler: Optional[ViewPooler]
@@ -266,6 +271,13 @@ class GenericModel(ImplicitronModelBase, torch.nn.Module):  # pyre-ignore: 13
         super().__init__()
         self.view_metrics = ViewMetrics()
 
+        if self.view_pooler_enabled:
+            if self.image_feature_extractor_class_type is None:
+                raise ValueError(
+                    "image_feature_extractor must be present for view pooling."
+                )
+        else:
+            self.image_feature_extractor_class_type = None
         run_auto_creation(self)
 
         self._implicit_functions = self._construct_implicit_functions()
@@ -349,20 +361,16 @@ class GenericModel(ImplicitronModelBase, torch.nn.Module):  # pyre-ignore: 13
         # custom_args hold additional arguments to the implicit function.
         custom_args = {}
 
-        if self.image_feature_extractor_enabled:
+        if self.image_feature_extractor is not None:
             # (2) Extract features for the image
-            img_feats = self.image_feature_extractor(  # pyre-fixme[29]
+            img_feats = self.image_feature_extractor(image_rgb, fg_probability)
-                image_rgb, fg_probability
+        else:
-            )
+            img_feats = None
 
         if self.view_pooler_enabled:
             if sequence_name is None:
                 raise ValueError("sequence_name must be provided for view pooling")
-            if not self.image_feature_extractor_enabled:
+            assert img_feats is not None
-                raise ValueError(
-                    "image_feature_extractor has to be enabled for for view pooling"
-                    + " (I.e. set self.image_feature_extractor_enabled=True)."
-                )
 
             # (3-4) Sample features and masks at the ray points.
             #       Aggregate features from multiple views.
@@ -555,13 +563,12 @@ class GenericModel(ImplicitronModelBase, torch.nn.Module):  # pyre-ignore: 13
                 **kwargs,
             )
 
-    def _get_viewpooled_feature_dim(self):
+    def _get_viewpooled_feature_dim(self) -> int:
-        return (
+        if self.view_pooler is None:
-            self.view_pooler.get_aggregated_feature_dim(
+            return 0
-                self.image_feature_extractor.get_feat_dims()
+        assert self.image_feature_extractor is not None
-            )
+        return self.view_pooler.get_aggregated_feature_dim(
-            if self.view_pooler_enabled
+            self.image_feature_extractor.get_feat_dims()
-            else 0
         )
 
     def create_raysampler(self):
@@ -617,11 +624,6 @@ class GenericModel(ImplicitronModelBase, torch.nn.Module):  # pyre-ignore: 13
         that image_feature_extractor is enabled when view_pooler is enabled.
         """
         if self.view_pooler_enabled:
-            if not self.image_feature_extractor_enabled:
-                raise ValueError(
-                    "image_feature_extractor has to be enabled for view pooling"
-                    + " (I.e. set self.image_feature_extractor_enabled=True)."
-                )
             self.view_pooler = ViewPooler(**self.view_pooler_args)
         else:
             self.view_pooler = None

tests/implicitron/data/overrides.yaml

@@ -17,7 +17,7 @@ sampling_mode_training: mask_sample
 sampling_mode_evaluation: full_grid
 raysampler_class_type: AdaptiveRaySampler
 renderer_class_type: LSTMRenderer
-image_feature_extractor_enabled: true
+image_feature_extractor_class_type: ResNetFeatureExtractor
 view_pooler_enabled: true
 implicit_function_class_type: IdrFeatureField
 loss_weights:
@@ -73,7 +73,7 @@ renderer_LSTMRenderer_args:
   hidden_size: 16
   n_feature_channels: 256
   verbose: false
-image_feature_extractor_args:
+image_feature_extractor_ResNetFeatureExtractor_args:
   name: resnet34
   pretrained: true
   stages:

tests/implicitron/test_config_use.py

@@ -9,6 +9,9 @@ import unittest
 
 from omegaconf import OmegaConf
 from pytorch3d.implicitron.models.autodecoder import Autodecoder
+from pytorch3d.implicitron.models.feature_extractor.resnet_feature_extractor import (
+    ResNetFeatureExtractor,
+)
 from pytorch3d.implicitron.models.generic_model import GenericModel
 from pytorch3d.implicitron.models.implicit_function.idr_feature_field import (
     IdrFeatureField,
@@ -63,7 +66,6 @@ class TestGenericModel(unittest.TestCase):
         provide_resnet34()
         args = get_default_args(GenericModel)
         args.view_pooler_enabled = True
-        args.image_feature_extractor_enabled = True
         args.view_pooler_args.feature_aggregator_class_type = (
             "AngleWeightedIdentityFeatureAggregator"
         )
@@ -77,9 +79,13 @@ class TestGenericModel(unittest.TestCase):
         )
         self.assertIsInstance(gm._implicit_functions[0]._fn, IdrFeatureField)
         self.assertIsInstance(gm.sequence_autodecoder, Autodecoder)
+        self.assertIsInstance(gm.image_feature_extractor, ResNetFeatureExtractor)
         self.assertFalse(hasattr(gm, "implicit_function"))
 
         instance_args = OmegaConf.structured(gm)
+        if DEBUG:
+            full_yaml = OmegaConf.to_yaml(instance_args, sort_keys=False)
+            (DATA_DIR / "overrides_full.yaml").write_text(full_yaml)
         remove_unused_components(instance_args)
         yaml = OmegaConf.to_yaml(instance_args, sort_keys=False)
         if DEBUG:

tests/implicitron/test_forward_pass.py

@@ -33,7 +33,7 @@ class TestGenericModel(unittest.TestCase):
         # Simple test of a forward and backward pass of the default GenericModel.
         device = torch.device("cuda:1")
         expand_args_fields(GenericModel)
-        model = GenericModel()
+        model = GenericModel(render_image_height=80, render_image_width=80)
         model.to(device)
         self._one_model_test(model, device)
 
@@ -149,6 +149,37 @@ class TestGenericModel(unittest.TestCase):
         )
         self.assertGreater(train_preds["objective"].item(), 0)
 
+    def test_viewpool(self):
+        device = torch.device("cuda:1")
+        args = get_default_args(GenericModel)
+        args.view_pooler_enabled = True
+        args.image_feature_extractor_ResNetFeatureExtractor_args.add_masks = False
+        model = GenericModel(**args)
+        model.to(device)
+
+        n_train_cameras = 2
+        R, T = look_at_view_transform(azim=torch.rand(n_train_cameras) * 360)
+        cameras = PerspectiveCameras(R=R, T=T, device=device)
+
+        defaulted_args = {
+            "fg_probability": None,
+            "depth_map": None,
+            "mask_crop": None,
+        }
+
+        target_image_rgb = torch.rand(
+            (n_train_cameras, 3, model.render_image_height, model.render_image_width),
+            device=device,
+        )
+        train_preds = model(
+            camera=cameras,
+            evaluation_mode=EvaluationMode.TRAINING,
+            image_rgb=target_image_rgb,
+            sequence_name=["a"] * n_train_cameras,
+            **defaulted_args,
+        )
+        self.assertGreater(train_preds["objective"].item(), 0)
+
 
 def _random_input_tensor(
     N: int,