torchvision 0.2.0 → 0.2.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 9aa604602112403b7f6738a7bb014812deb528b9a4d480ba89cf7f0c6d01b59c
-  data.tar.gz: 3ffa29d3ff5234040df51113d085a807d94a45c6998f8182db72bf60c25d67e7
+  metadata.gz: bbb87c59c0f081c0de57ccdd62e30bfc551e1cb69523e4ffd498c997e1a2d8b3
+  data.tar.gz: 890da113706e659d57194980c5c9262075beb8398a75da2997c0812b70abe308
 SHA512:
-  metadata.gz: 89f61279aed314c84ae33c14efe266b849bf74a40a9f90beb650358610b479cb44f6cd247ff56aa89d0227ef6358973b5055fd1ce9bcf1e973402b2118bfe75d
-  data.tar.gz: 4eb63bfd9a79bd3683238c186e0e70aba494028f54ce8f971a551f362eec4b4e55fe83c8e738c7669fde2c3671e3360a01b02c4d29f6c965640a6c08b3c4379d
+  metadata.gz: 3445b62b7824ae16205034881d37c48ac4c70d7e5677014755ae5600632f9ce45168f41b0d3e98c8104eb8337e1566db4df3e0ad5ace5e6a46a5d213d01b6c8d
+  data.tar.gz: 93f22c385586ff8a010880676806f6bc9ba2f614c4c14886235a300ea6e2abce0f80c260e255644e1b4d24e6ecddfd21830dde2960a92a9492239e69622d4548

data/CHANGELOG.md

@@ -1,3 +1,9 @@
+## 0.2.1 (2021-03-14)
+
+- Added `ImageFolder` and `DatasetFolder`
+- Added `CenterCrop` and `RandomResizedCrop` transforms
+- Added `crop` method
+
 ## 0.2.0 (2021-03-11)
 
 - Added `RandomHorizontalFlip`, `RandomVerticalFlip`, and `Resize` transforms

data/README.md

@@ -6,7 +6,7 @@
 
 ## Installation
 
-First, [install libvips](libvips-installation). For Homebrew, use:
+First, [install libvips](#libvips-installation). For Homebrew, use:
 
 ```sh
 brew install vips
@@ -25,7 +25,8 @@ This library follows the [Python API](https://pytorch.org/docs/stable/torchvisio
 ## Examples
 
 - [MNIST](https://github.com/ankane/torch.rb/tree/master/examples/mnist)
-- [Generative Adversarial Networks](https://github.com/ankane/torch.rb/tree/master/examples/gan)
+- [Transfer learning](https://github.com/ankane/torch.rb/tree/master/examples/transfer-learning)
+- [Generative adversarial networks](https://github.com/ankane/torch.rb/tree/master/examples/gan)
 
 ## Datasets
 
@@ -54,9 +55,11 @@ TorchVision::Transforms::Compose.new([
 
 Supported transforms are:
 
+- CenterCrop
 - Compose
 - Normalize
 - RandomHorizontalFlip
+- RandomResizedCrop
 - RandomVerticalFlip
 - Resize
 - ToTensor
@@ -130,12 +133,16 @@ net.load_state_dict(Torch.load("net.pth"))
 
 ## libvips Installation
 
-### Ubuntu
+### Linux
+
+Check your package manager. For Ubuntu, use:
 
 ```sh
 sudo apt install libvips
 ```
 
+You can also [build from source](https://libvips.github.io/libvips/install.html).
+
 ### Mac
 
 ```sh

data/lib/torchvision.rb

@@ -16,6 +16,8 @@ require "torchvision/version"
 
 # datasets
 require "torchvision/datasets/vision_dataset"
+require "torchvision/datasets/dataset_folder"
+require "torchvision/datasets/image_folder"
 require "torchvision/datasets/cifar10"
 require "torchvision/datasets/cifar100"
 require "torchvision/datasets/mnist"
@@ -47,10 +49,12 @@ require "torchvision/models/wide_resnet50_2"
 require "torchvision/models/wide_resnet101_2"
 
 # transforms
+require "torchvision/transforms/center_crop"
 require "torchvision/transforms/compose"
 require "torchvision/transforms/functional"
 require "torchvision/transforms/normalize"
 require "torchvision/transforms/random_horizontal_flip"
+require "torchvision/transforms/random_resized_crop"
 require "torchvision/transforms/random_vertical_flip"
 require "torchvision/transforms/resize"
 require "torchvision/transforms/to_tensor"

data/lib/torchvision/datasets/dataset_folder.rb

@@ -0,0 +1,91 @@
+module TorchVision
+  module Datasets
+    class DatasetFolder < VisionDataset
+      attr_reader :classes
+
+      def initialize(root, extensions: nil, transform: nil, target_transform: nil, is_valid_file: nil)
+        super(root, transform: transform, target_transform: target_transform)
+        classes, class_to_idx = find_classes(@root)
+        samples = make_dataset(@root, class_to_idx, extensions, is_valid_file)
+        if samples.empty?
+          msg = "Found 0 files in subfolders of: #{@root}\n"
+          unless extensions.nil?
+            msg += "Supported extensions are: #{extensions.join(",")}"
+          end
+          raise RuntimeError, msg
+        end
+
+        @loader = lambda do |path|
+          Vips::Image.new_from_file(path)
+        end
+        @extensions = extensions
+
+        @classes = classes
+        @class_to_idx = class_to_idx
+        @samples = samples
+        @targets = samples.map { |s| s[1] }
+      end
+
+      def [](index)
+        path, target = @samples[index]
+        sample = @loader.call(path)
+        if @transform
+          sample = @transform.call(sample)
+        end
+        if @target_transform
+          target = @target_transform.call(target)
+        end
+
+        [sample, target]
+      end
+
+      def size
+        @samples.size
+      end
+
+      private
+
+      def find_classes(dir)
+        classes = Dir.children(dir).select { |d| File.directory?(File.join(dir, d)) }
+        classes.sort!
+        class_to_idx = classes.map.with_index.to_h
+        [classes, class_to_idx]
+      end
+
+      def has_file_allowed_extension(filename, extensions)
+        filename = filename.downcase
+        extensions.any? { |ext| filename.end_with?(ext) }
+      end
+
+      def make_dataset(directory, class_to_idx, extensions, is_valid_file)
+        instances = []
+        directory = File.expand_path(directory)
+        both_none = extensions.nil? && is_valid_file.nil?
+        both_something = !extensions.nil? && !is_valid_file.nil?
+        if both_none || both_something
+          raise ArgumentError, "Both extensions and is_valid_file cannot be None or not None at the same time"
+        end
+        if !extensions.nil?
+          is_valid_file = lambda do |x|
+            has_file_allowed_extension(x, extensions)
+          end
+        end
+        class_to_idx.keys.sort.each do |target_class|
+          class_index = class_to_idx[target_class]
+          target_dir = File.join(directory, target_class)
+          if !File.directory?(target_dir)
+            next
+          end
+          Dir.glob("**", base: target_dir).sort.each do |fname|
+            path = File.join(target_dir, fname)
+            if is_valid_file.call(path)
+              item = [path, class_index]
+              instances << item
+            end
+          end
+        end
+        instances
+      end
+    end
+  end
+end

data/lib/torchvision/datasets/image_folder.rb

@@ -0,0 +1,12 @@
+module TorchVision
+  module Datasets
+    class ImageFolder < DatasetFolder
+      IMG_EXTENSIONS = [".jpg", ".jpeg", ".png", ".ppm", ".bmp", ".pgm", ".tif", ".tiff", ".webp"]
+
+      def initialize(root, transform: nil, target_transform: nil, is_valid_file: nil)
+        super(root, extensions: IMG_EXTENSIONS, transform: transform, target_transform: target_transform, is_valid_file: is_valid_file)
+        @imgs = @samples
+      end
+    end
+  end
+end

data/lib/torchvision/datasets/vision_dataset.rb

@@ -1,7 +1,6 @@
 module TorchVision
   module Datasets
-    # TODO inherit Torch::Utils::Data::Dataset
-    class VisionDataset
+    class VisionDataset < Torch::Utils::Data::Dataset
       attr_reader :data, :targets
 
       def initialize(root, transforms: nil, transform: nil, target_transform: nil)

data/lib/torchvision/transforms/center_crop.rb

@@ -0,0 +1,13 @@
+module TorchVision
+  module Transforms
+    class CenterCrop < Torch::NN::Module
+      def initialize(size)
+        @size = size
+      end
+
+      def forward(img)
+        F.center_crop(img, @size)
+      end
+    end
+  end
+end

data/lib/torchvision/transforms/functional.rb

@@ -34,21 +34,23 @@ module TorchVision
 
         def resize(img, size)
           raise "img should be Vips::Image. Got #{img.class.name}" unless img.is_a?(Vips::Image)
-          # TODO support array size
-          raise "Got inappropriate size arg: #{size}" unless size.is_a?(Integer)
 
-          w, h = img.size
-          if (w <= h && w == size) || (h <= w && h == size)
-            return img
-          end
-          if w < h
-            ow = size
-            oh = (size * h / w).to_i
-            img.thumbnail_image(ow, height: oh)
+          if size.is_a?(Integer)
+            w, h = img.size
+            if (w <= h && w == size) || (h <= w && h == size)
+              return img
+            end
+            if w < h
+              ow = size
+              oh = (size * h / w).to_i
+              img.thumbnail_image(ow, height: oh)
+            else
+              oh = size
+              ow = (size * w / h).to_i
+              img.thumbnail_image(ow, height: oh)
+            end
           else
-            oh = size
-            ow = (size * w / h).to_i
-            img.thumbnail_image(ow, height: oh)
+            img.thumbnail_image(size[0], height: size[1], size: :force)
           end
         end
 
@@ -90,6 +92,7 @@ module TorchVision
 
         def hflip(img)
           if img.is_a?(Torch::Tensor)
+            assert_image_tensor(img)
             img.flip(-1)
           else
             img.flip(:horizontal)
@@ -98,11 +101,76 @@ module TorchVision
 
         def vflip(img)
           if img.is_a?(Torch::Tensor)
+            assert_image_tensor(img)
             img.flip(-2)
           else
             img.flip(:vertical)
           end
         end
+
+        def crop(img, top, left, height, width)
+          if img.is_a?(Torch::Tensor)
+            assert_image_tensor(img)
+            indexes = [true] * (img.dim - 2)
+            img[*indexes, top...(top + height), left...(left + width)]
+          else
+            img.crop(left, top, width, height)
+          end
+        end
+
+        def center_crop(img, output_size)
+          if output_size.is_a?(Integer)
+            output_size = [output_size.to_i, output_size.to_i]
+          elsif output_size.is_a?(Array) && output_size.length == 1
+            output_size = [output_size[0], output_size[0]]
+          end
+
+          image_width, image_height = image_size(img)
+          crop_height, crop_width = output_size
+
+          if crop_width > image_width || crop_height > image_height
+            padding_ltrb = [
+              crop_width > image_width ? (crop_width - image_width).div(2) : 0,
+              crop_height > image_height ? (crop_height - image_height).div(2) : 0,
+              crop_width > image_width ? (crop_width - image_width + 1).div(2) : 0,
+              crop_height > image_height ? (crop_height - image_height + 1).div(2) : 0
+            ]
+            # TODO
+            img = pad(img, padding_ltrb, fill: 0)
+            image_width, image_height = image_size(img)
+            if crop_width == image_width && crop_height == image_height
+              return img
+            end
+          end
+
+          crop_top = ((image_height - crop_height) / 2.0).round
+          crop_left = ((image_width - crop_width) / 2.0).round
+          crop(img, crop_top, crop_left, crop_height, crop_width)
+        end
+
+        # TODO interpolation
+        def resized_crop(img, top, left, height, width, size)
+          img = crop(img, top, left, height, width)
+          img = resize(img, size) #, interpolation)
+          img
+        end
+
+        private
+
+        def image_size(img)
+          if img.is_a?(Torch::Tensor)
+            assert_image_tensor(img)
+            [img.shape[-1], img.shape[-2]]
+          else
+            [img.width, img.height]
+          end
+        end
+
+        def assert_image_tensor(img)
+          if img.ndim < 2
+            raise TypeError, "Tensor is not a torch image."
+          end
+        end
       end
     end
 

data/lib/torchvision/transforms/random_resized_crop.rb

@@ -0,0 +1,70 @@
+module TorchVision
+  module Transforms
+    class RandomResizedCrop < Torch::NN::Module
+      def initialize(size, scale: [0.08, 1.0], ratio: [3.0 / 4.0, 4.0 / 3.0])
+        super()
+        @size = setup_size(size, "Please provide only two dimensions (h, w) for size.")
+        # @interpolation = interpolation
+        @scale = scale
+        @ratio = ratio
+      end
+
+      def params(img, scale, ratio)
+        width, height = F.send(:image_size, img)
+        area = height * width
+
+        log_ratio = Torch.log(Torch.tensor(ratio))
+        10.times do
+          target_area = area * Torch.empty(1).uniform!(scale[0], scale[1]).item
+          aspect_ratio = Torch.exp(
+            Torch.empty(1).uniform!(log_ratio[0], log_ratio[1])
+          ).item
+
+          w = Math.sqrt(target_area * aspect_ratio).round
+          h = Math.sqrt(target_area / aspect_ratio).round
+
+          if 0 < w && w <= width && 0 < h && h <= height
+            i = Torch.randint(0, height - h + 1, size: [1]).item
+            j = Torch.randint(0, width - w + 1, size: [1]).item
+            return i, j, h, w
+          end
+        end
+
+        # Fallback to central crop
+        in_ratio = width.to_f / height.to_f
+        if in_ratio < ratio.min
+          w = width
+          h = (w / ratio.min).round
+        elsif in_ratio > ratio.max
+          h = height
+          w = (h * ratio.max).round
+        else # whole image
+          w = width
+          h = height
+        end
+        i = (height - h).div(2)
+        j = (width - w).div(2)
+        [i, j, h, w]
+      end
+
+      def forward(img)
+        i, j, h, w = params(img, @scale, @ratio)
+        F.resized_crop(img, i, j, h, w, @size) #, @interpolation)
+      end
+
+      private
+
+      def setup_size(size, error_msg)
+        if size.is_a?(Integer)
+          return [size, size]
+        end
+
+        if size.length != 2
+          raise ArgumentError, error_msg
+        end
+
+        size
+      end
+    end
+  end
+end

data/lib/torchvision/version.rb

@@ -1,3 +1,3 @@
 module TorchVision
-  VERSION = "0.2.0"
+  VERSION = "0.2.1"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: torchvision
 version: !ruby/object:Gem::Version
-  version: 0.2.0
+  version: 0.2.1
 platform: ruby
 authors:
 - Andrew Kane
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2021-03-11 00:00:00.000000000 Z
+date: 2021-03-15 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: numo-narray
@@ -64,7 +64,9 @@ files:
 - lib/torchvision.rb
 - lib/torchvision/datasets/cifar10.rb
 - lib/torchvision/datasets/cifar100.rb
+- lib/torchvision/datasets/dataset_folder.rb
 - lib/torchvision/datasets/fashion_mnist.rb
+- lib/torchvision/datasets/image_folder.rb
 - lib/torchvision/datasets/kmnist.rb
 - lib/torchvision/datasets/mnist.rb
 - lib/torchvision/datasets/vision_dataset.rb
@@ -90,10 +92,12 @@ files:
 - lib/torchvision/models/vgg19_bn.rb
 - lib/torchvision/models/wide_resnet101_2.rb
 - lib/torchvision/models/wide_resnet50_2.rb
+- lib/torchvision/transforms/center_crop.rb
 - lib/torchvision/transforms/compose.rb
 - lib/torchvision/transforms/functional.rb
 - lib/torchvision/transforms/normalize.rb
 - lib/torchvision/transforms/random_horizontal_flip.rb
+- lib/torchvision/transforms/random_resized_crop.rb
 - lib/torchvision/transforms/random_vertical_flip.rb
 - lib/torchvision/transforms/resize.rb
 - lib/torchvision/transforms/to_tensor.rb