Update DataAugmentations dependency, fix docs (#17)

* Minor fixes to documentation

* Update DataAugmentations dependency

* Fix test, update CI

* Update DA test refs

Author: adrhill

.github/workflows/CI.yml

@@ -24,7 +24,8 @@ jobs:
       matrix:
         version:
           - '1'
-          - 'nightly'
+          - 'lts'
+          - 'pre'
         os:
           - ubuntu-latest
         arch:

Project.toml

@@ -1,7 +1,7 @@
 name = "ImageNetDataset"
 uuid = "d141d81c-f7c4-4296-8313-8866e6f9f78a"
 authors = ["Adrian Hill <[email protected]>"]
-version = "1.0.0-DEV"
+version = "0.1.0"
 
 [deps]
 DataAugmentation = "88a5189c-e7ff-4f85-ac6b-e6158070f02e"
@@ -19,7 +19,7 @@ DataAugmentation = "88a5189c-e7ff-4f85-ac6b-e6158070f02e"
 ImageNetDatasetDataAugmentationExt = "DataAugmentation"
 
 [compat]
-DataAugmentation = "0.2"
+DataAugmentation = "0.3"
 DataDeps = "0.7"
 ImageCore = "0.10"
 JpegTurbo = "0.1"

README.md

@@ -58,7 +58,7 @@ dataset = ImageNet(:val; transform=tfm)
 Custom transformations can be implemented by extending `AbstractTransformation`.
 
 To apply a transformation outside of the `ImageNet` dataset,
-e.g. to preprocess a single image add a given `path`, run
+e.g. to preprocess a single image at a given `path`, run
 
 ```julia
 transform(tfm, path)
@@ -82,6 +82,6 @@ dataset = ImageNet(:val; transform=tfm)
 
 > [!NOTE]
 > This repository is based on [MLDatasets.jl PR #146](https://github.com/JuliaML/MLDatasets.jl/pull/146)
-> and mirrors the MLDatasets API.
+> and mirrors the MLDatasets `v0.7` API.
 >
 > Copyright (c) 2015 Hiroyuki Shindo and contributors.

docs/src/installation.md

@@ -70,23 +70,25 @@ Download the following files from the [ILSVRC2012 download page](https://image-n
 You can use ImageNetDataset.jl without downloading all splits.
 
 After downloading the data, move and extract the training and validation images to
-labeled subfolders by running the following shell script:
+labeled subfolders by running the following shell scripts:
 
+**Extract metadata from the devkit:**
+```bash
+mkdir -p ImageNet/devkit && tar -xvf ILSVRC2012_devkit_t12.tar -C ImageNet/devkit --strip-components=1
+```
+
+**Extract the training data:**
 ```bash
-# Extract the training data:
 mkdir -p ImageNet/train && tar -xvf ILSVRC2012_img_train.tar -C ImageNet/train
 
 # Unpack all 1000 compressed tar-files, one for each category:
 cd ImageNet/train
 find . -name "*.tar" | while read NAME ; do mkdir -p "\${NAME%.tar}"; tar -xvf "\${NAME}" -C "\${NAME%.tar}"; rm -f "\${NAME}"; done
+```
 
-# Extract the validation data:
-cd ../..
+**Extract the validation data:**
+```bash
 mkdir -p ImageNet/val && tar -xvf ILSVRC2012_img_val.tar -C ImageNet/val
-
-# Extract metadata from the devkit:
-cd ../..
-mkdir -p ImageNet/devkit && tar -xvf ILSVRC2012_devkit_t12.tar.gz -C ImageNet/devkit --strip-components=1
 ```
 
 And run the following script 
@@ -97,4 +99,3 @@ to create all class directories and move images into corresponding directories:
 cd ImageNet/val
 wget -qO- https://raw.githubusercontent.com/Julia-XAI/ImageNetDataset.jl/master/docs/src/valprep.sh | bash
 ```
-    

src/imagenet.jl

@@ -131,8 +131,6 @@ function ImageNet(
     test_dir="test",
     devkit_dir="devkit",
 )
-    @assert split ∈ (:train, :val, :test)
-
     depname = "ImageNet"
     root_dir = get_datadep_dir(depname, dir)
 
@@ -146,9 +144,11 @@ function ImageNet(
     elseif split == :val
         paths = get_file_paths(joinpath(root_dir, val_dir))
         @assert length(paths) == VALSET_SIZE
-    else
+    elseif split == :test
         paths = get_file_paths(joinpath(root_dir, test_dir))
         @assert length(paths) == TESTSET_SIZE
+    else
+        throw(ArgumentError("Unknown split :$split"))
     end
     targets = [metadata["wnid_to_label"][wnid] for wnid in get_wnids(paths)]
     return ImageNet(split, transform, paths, targets, metadata)

test/references/DataAugmentations.txt

@@ -43,7 +43,7 @@ img = inverse_transform(tfm, A)
 @test size(img) == (40, 60)
 
 # DataAugmentations.jl
-tfm = CenterResizeCrop((60, 40)) |> ImageToTensor() |> Normalize(default_mean, default_std)
+tfm = CenterResizeCrop((40, 60)) |> ImageToTensor() |> Normalize(default_mean, default_std)
 A = transform(tfm, path)
 @test size(A) == (60, 40, 3)
 @test_reference "references/DataAugmentations.txt" A