Layers branch with NMS-GPU

.coveragerc

@@ -0,0 +1,7 @@
+[run]
+branch = True
+
+[paths]
+source = 
+    torchvision
+    /**/site-packages/torchvision

.gitignore

@@ -6,3 +6,5 @@ torchvision.egg-info/
 */**/*~
 *~
 docs/build
+.coverage
+htmlcov

.travis.yml

@@ -6,14 +6,15 @@ matrix:
       python: "2.7"
       install: pip install flake8
       script: flake8
+      after_success: []
     - python: "2.7"
       env: IMAGE_BACKEND=Pillow-SIMD
     - python: "2.7"
     - python: "3.5"
       env: IMAGE_BACKEND=Pillow-SIMD
     - python: "3.5"
 
-install:
+before_install:
   - sudo apt-get update
   - wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh -O miniconda.sh;
   - bash miniconda.sh -b -p $HOME/miniconda
@@ -26,11 +27,34 @@ install:
 
   - conda create -q -n test-environment python=$TRAVIS_PYTHON_VERSION pytorch scipy -c pytorch
   - source activate test-environment
-  - python setup.py install
-  - pip install --upgrade pytest
-  - if [[ "$IMAGE_BACKEND" == "Pillow-SIMD" ]]; then
-      pip uninstall -y pillow && CC="cc -march=native" pip install --force-reinstall pillow-simd;
+  - |
+    if [[ "$IMAGE_BACKEND" == "Pillow-SIMD" ]]; then
+      pip uninstall -y pillow && CC="cc -march=native" pip install --force-reinstall pillow-simd
     fi
+  - pip install pytest pytest-cov codecov
+
+
+install:
+  # Using pip instead of setup.py ensures we install a non-compressed version of the package
+  # (as opposed to an egg), which is necessary to collect coverage.
+  # We still get the benefit of testing an installed version over the
+  # test version to iron out installation file-inclusion bugs but can
+  # also collect coverage.
+  - pip install .
+  # Move to home dir, otherwise we'll end up with the path to the
+  # package in $PWD rather than the installed v
+  - |
+    cd $HOME
+    export TV_INSTALL_PATH="$(python -c 'import os; import torchvision; print(os.path.dirname(os.path.abspath(torchvision.__file__)))')"
+    echo "$TV_INSTALL_PATH"
+    cd -
+
 script:
-  - pytest test/
+  - pytest --cov-config .coveragerc --cov torchvision --cov $TV_INSTALL_PATH test
 
+after_success:
+  # Necessary to run coverage combine to rewrite paths from
+  # /travis/env/path/site-packages/torchvision to actual path
+  - coverage combine .coverage
+  - coverage report
+  - codecov

README.rst

@@ -4,6 +4,16 @@ torchvision
 .. image:: https://travis-ci.org/pytorch/vision.svg?branch=master
     :target: https://travis-ci.org/pytorch/vision
 
+.. image:: https://codecov.io/gh/pytorch/vision/branch/master/graph/badge.svg
+    :target: https://codecov.io/gh/pytorch/vision
+
+.. image:: https://pepy.tech/badge/torchvision
+    :target: https://pepy.tech/project/torchvision
+
+.. image:: https://img.shields.io/badge/dynamic/json.svg?label=docs&url=https%3A%2F%2Fpypi.org%2Fpypi%2Ftorchvision%2Fjson&query=%24.info.version&colorB=brightgreen&prefix=v
+    :target: https://pytorch.org/docs/stable/torchvision/index.html
+
+
 The torchvision package consists of popular datasets, model architectures, and common image transformations for computer vision.
 
 Installation

docs/source/datasets.rst

@@ -35,6 +35,11 @@ Fashion-MNIST
 
 .. autoclass:: FashionMNIST
 
+KMNIST
+~~~~~~~~~~~~~
+
+.. autoclass:: KMNIST
+
 EMNIST
 ~~~~~~
 
@@ -129,3 +134,45 @@ PhotoTour
 .. autoclass:: PhotoTour
   :members: __getitem__
   :special-members:
+
+SBU
+~~~
+
+
+.. autoclass:: SBU
+  :members: __getitem__
+  :special-members:
+
+Flickr
+~~~~~~
+
+
+.. autoclass:: Flickr8k
+  :members: __getitem__
+  :special-members:
+
+.. autoclass:: Flickr30k
+  :members: __getitem__
+  :special-members:
+
+VOC
+~~~~~~
+
+
+.. autoclass:: VOCSegmentation
+  :members: __getitem__
+  :special-members:
+
+.. autoclass:: VOCDetection
+  :members: __getitem__
+  :special-members:
+
+Cityscapes
+~~~~~~~~~~
+
+.. note ::
+    Requires Cityscape to be downloaded.
+
+.. autoclass:: Cityscapes
+  :members: __getitem__
+  :special-members:

docs/source/models.rst

@@ -36,6 +36,10 @@ These can be constructed by passing ``pretrained=True``:
     densenet = models.densenet161(pretrained=True)
     inception = models.inception_v3(pretrained=True)
 
+Instancing a pre-trained model will download its weights to a cache directory.
+This directory can be set using the `TORCH_MODEL_ZOO` environment variable. See
+:func:`torch.utils.model_zoo.load_url` for details.
+
 Some models use modules which have different training and evaluation
 behavior, such as batch normalization. To switch between these modes, use
 ``model.train()`` or ``model.eval()`` as appropriate. See

docs/source/transforms.rst

@@ -3,61 +3,72 @@ torchvision.transforms
 
 .. currentmodule:: torchvision.transforms
 
-Transforms are common image transforms. They can be chained together using :class:`Compose`
+Transforms are common image transformations. They can be chained together using :class:`Compose`.
+Additionally, there is the :mod:`torchvision.transforms.functional` module.
+Functional transforms give fine-grained control over the transformations.
+This is useful if you have to build a more complex transformation pipeline
+(e.g. in the case of segmentation tasks).
 
 .. autoclass:: Compose
 
 Transforms on PIL Image
 -----------------------
 
-.. autoclass:: Resize
+.. autoclass:: CenterCrop
 
-.. autoclass:: Scale
+.. autoclass:: ColorJitter
 
-.. autoclass:: CenterCrop
+.. autoclass:: FiveCrop
 
-.. autoclass:: RandomCrop
+.. autoclass:: Grayscale
 
-.. autoclass:: RandomHorizontalFlip
+.. autoclass:: Pad
 
-.. autoclass:: RandomVerticalFlip
+.. autoclass:: RandomAffine
 
-.. autoclass:: RandomResizedCrop
+.. autoclass:: RandomApply
 
-.. autoclass:: RandomSizedCrop
+.. autoclass:: RandomChoice
 
-.. autoclass:: Grayscale
+.. autoclass:: RandomCrop
 
 .. autoclass:: RandomGrayscale
 
-.. autoclass:: FiveCrop
-
-.. autoclass:: TenCrop
+.. autoclass:: RandomHorizontalFlip
 
-.. autoclass:: Pad
+.. autoclass:: RandomOrder
 
-.. autoclass:: ColorJitter
+.. autoclass:: RandomResizedCrop
 
 .. autoclass:: RandomRotation
 
-.. autoclass:: RandomAffine
+.. autoclass:: RandomSizedCrop
+
+.. autoclass:: RandomVerticalFlip
+
+.. autoclass:: Resize
+
+.. autoclass:: Scale
+
+.. autoclass:: TenCrop
 
 Transforms on torch.\*Tensor
 ----------------------------
 
+.. autoclass:: LinearTransformation
+
 .. autoclass:: Normalize
 	:members: __call__
 	:special-members:
 
-
 Conversion Transforms
 ---------------------
 
-.. autoclass:: ToTensor
+.. autoclass:: ToPILImage
 	:members: __call__
 	:special-members:
 
-.. autoclass:: ToPILImage
+.. autoclass:: ToTensor
 	:members: __call__
 	:special-members:
 
@@ -66,3 +77,28 @@ Generic Transforms
 
 .. autoclass:: Lambda
 
+
+Functional Transforms
+---------------------
+
+Functional transforms give you fine-grained control of the transformation pipeline.
+As opposed to the transformations above, functional transforms don't contain a random number
+generator for their parameters.
+That means you have to specify/generate all parameters, but you can reuse the functional transform.
+For example, you can apply a functional transform to multiple images like this:
+
+.. code:: python
+
+    import torchvision.transforms.functional as TF
+    import random
+
+    def my_segmentation_transforms(image, segmentation):
+        if random.random() > 5:
+            angle = random.randint(-30, 30)
+            image = TF.rotate(image, angle)
+            segmentation = TF.rotate(segmentation, angle)
+        # more transforms ...
+        return image, segmentation
+
+.. automodule:: torchvision.transforms.functional
+    :members:

hubconf.py

@@ -0,0 +1,65 @@
+'''
+This is an example hubconf.py for pytorch/vision repo
+
+## Users can get this published model by calling:
+hub_model = hub.load(
+    'pytorch/vision:master', # repo_owner/repo_name:branch
+    'resnet18', # entrypoint
+    1234, # args for callable [not applicable to resnet]
+    pretrained=True) # kwargs for callable
+
+## Protocol on repo owner side
+1. The "published" models should be at least in a branch/tag. It can't be a random commit.
+2. Repo owner should have the following field defined in hubconf.py
+  2.1 Function/entrypoint with function signature "def resnet18(pretrained=False, *args, **kwargs):"
+  2.2 Pretrained allows users to load pretrained weights from repo owner.
+  2.3 Args and kwargs are passed to the callable _resnet18,
+  2.4 Docstring of function works as a help message, explaining what does the model do and what's
+      the allowed arguments.
+  2.5 Dependencies is a list optionally provided by repo owner, to specify what packages are required
+      to run the model.
+
+## Hub_dir
+
+hub_dir specifies where the intermediate files/folders will be saved. By default this is ~/.torch/hub.
+Users can change it by either setting the environment variable TORCH_HUB_DIR or calling hub.set_dir(PATH_TO_HUB_DIR).
+By default, we don't cleanup files after loading so that users can use cache next time.
+
+## Cache logic
+
+We used the cache by default if it exists in hub_dir.
+Users can force a fresh reload by calling hub.load(..., force_reload=True).
+'''
+
+import torch.utils.model_zoo as model_zoo
+
+# Optional list of dependencies required by the package
+dependencies = ['torch', 'math']
+
+
+def resnet18(pretrained=False, *args, **kwargs):
+    """
+    Resnet18 model
+    pretrained (bool): a recommended kwargs for all entrypoints
+    args & kwargs are arguments for the function
+    """
+    from torchvision.models.resnet import resnet18 as _resnet18
+    model = _resnet18(*args, **kwargs)
+    checkpoint = 'https://download.pytorch.org/models/resnet18-5c106cde.pth'
+    if pretrained:
+        model.load_state_dict(model_zoo.load_url(checkpoint, progress=False))
+    return model
+
+
+def resnet50(pretrained=False, *args, **kwargs):
+    """
+    Resnet50 model
+    pretrained (bool): a recommended kwargs for all entrypoints
+    args & kwargs are arguments for the function
+    """
+    from torchvision.models.resnet import resnet50 as _resnet50
+    model = _resnet50(*args, **kwargs)
+    checkpoint = 'https://download.pytorch.org/models/resnet50-19c8e357.pth'
+    if pretrained:
+        model.load_state_dict(model_zoo.load_url(checkpoint, progress=False))
+    return model

setup.cfg

@@ -9,5 +9,5 @@ max-line-length = 120
 
 [flake8]
 max-line-length = 120
-ignore = F401,E402,F403
+ignore = F401,E402,F403,W503,W504
 exclude = venv