Samples and unit test for image-related transform estimators #3165

abgoswam · 2019-04-02T01:30:46Z

Towards #1209

The PR makes the following changes

Adds unit test and sample for the ConvertToImage transform estimator.
Fixes the info presented to the user for the 4 existing image samples {ConvertToGrayscale, LoadImages, ExtractPixels, ResizeImages}

sfilipi · 2019-04-02T05:09:52Z

docs/samples/Microsoft.ML.Samples/Dynamic/Transforms/ImageAnalytics/ConvertToGrayScale.cs

+            // ImagePath : tomato.bmp
+            // Name : tomato
+            // ImageObject : System.Drawing.Bitmap
+            // Grayscale : System.Drawing.Bitmap


maybe transpose, to give the column/row idea more accurately. #Resolved

fixed

In reply to: 271135652 [](ancestors = 271135652)

sfilipi · 2019-04-02T05:10:00Z

docs/samples/Microsoft.ML.Samples/Dynamic/Transforms/ImageAnalytics/ConvertToGrayScale.cs

+            // Preview 1 row of the transformedData. 
+            var transformedDataPreview = transformedData.Preview(1);
+            foreach (var kvPair in transformedDataPreview.RowView[0].Values)
+            {


{ [](start = 12, length = 1)

omit :) #Resolved

sfilipi · 2019-04-02T05:10:23Z

docs/samples/Microsoft.ML.Samples/Dynamic/Transforms/ImageAnalytics/ConvertToImage.cs

+{
+    public static class ConvertToImage
+    {
+        private const int inputSize = 3 * 224 * 224;


224 * 224; [](start = 39, length = 13)

might want to add a comment about what does each number represent #Resolved

sfilipi · 2019-04-02T05:11:13Z

docs/samples/Microsoft.ML.Samples/Dynamic/Transforms/ImageAnalytics/ConvertToImage.cs

+                    Features = Enumerable.Repeat(0, inputSize).Select(x => random.NextDouble() * 100).ToArray()
+                };
+            }
+        }


i'd move this below the example. If the users want to scroll to it they can. #Resolved

fixed

In reply to: 271135840 [](ancestors = 271135840)

sfilipi · 2019-04-02T05:11:45Z

docs/samples/Microsoft.ML.Samples/Dynamic/Transforms/ImageAnalytics/ConvertToImage.cs

+            // Preview 1 row of the transformedData. 
+            var transformedDataPreview = transformedData.Preview(1);
+            foreach (var kvPair in transformedDataPreview.RowView[0].Values)
+            {


{ [](start = 11, length = 2)

omit #Resolved

sfilipi · 2019-04-02T05:12:05Z

docs/samples/Microsoft.ML.Samples/Dynamic/Transforms/ImageAnalytics/ConvertToImage.cs

+
+            // The transformedData IDataView contains the loaded and resized raw values
+
+            // Preview 1 row of the transformedData. 


no need for space in between. #Resolved

sfilipi · 2019-04-02T05:12:26Z

docs/samples/Microsoft.ML.Samples/Dynamic/Transforms/ImageAnalytics/ConvertToImage.cs

+            // Pixels : Dense vector of size 30000
+
+            Console.WriteLine("--------------------------------------------------");
+


no need for it. Users can add it #Resolved

sfilipi · 2019-04-02T05:13:24Z

docs/samples/Microsoft.ML.Samples/Dynamic/Transforms/ImageAnalytics/ConvertToImage.cs

+            Console.WriteLine("--------------------------------------------------");
+
+            // Using schema comprehension to display raw pixels for each row.
+            // Display extracted pixels in column 'Pixels'.


might be easier to understand if put as:

Convert the transformedData into an IEnumerable #Resolved

modified the pretty printing method

In reply to: 271136133 [](ancestors = 271136133)

sfilipi · 2019-04-02T05:14:21Z

docs/samples/Microsoft.ML.Samples/Dynamic/Transforms/ImageAnalytics/ExtractPixels.cs


-            // Preview of the content of the images.tsv file
-            //
-            // ImagePath    Name        ImageObject                 "Pixels"


// ImagePath Name ImageObject [](start = 11, length = 41)

maybe keep the headers/column names, even if they don't print out. Or print them separately. #Resolved

fixed

In reply to: 271136252 [](ancestors = 271136252)

Ivanidzo4ka · 2019-04-02T17:05:21Z

docs/samples/Microsoft.ML.Samples/Dynamic/Transforms/ImageAnalytics/LoadImages.cs


+            var transformedData = pipeline.Fit(data).Transform(data);
            // The transformedData IDataView contains the loaded images now


now [](start = 72, length = 3)

dot in the end of sentence :) #Resolved

Ivanidzo4ka · 2019-04-02T17:05:50Z

docs/samples/Microsoft.ML.Samples/Dynamic/Transforms/ImageAnalytics/ResizeImages.cs


            var transformedData = pipeline.Fit(data).Transform(data);
+            // The transformedData IDataView contains the resized images now


dot here is well #Resolved

Ivanidzo4ka

codecov · 2019-04-02T17:31:40Z

Codecov Report

Merging #3165 into master will increase coverage by 0.01%.
The diff coverage is 94.73%.

@@            Coverage Diff             @@
##           master    #3165      +/-   ##
==========================================
+ Coverage   72.53%   72.55%   +0.01%     
==========================================
  Files         808      807       -1     
  Lines      144775   144793      +18     
  Branches    16209    16211       +2     
==========================================
+ Hits       105012   105054      +42     
+ Misses      35348    35323      -25     
- Partials     4415     4416       +1

Flag	Coverage Δ
#Debug	`72.55% <94.73%> (+0.01%)`	⬆️
#production	`68.14% <ø> (+0.02%)`	⬆️
#test	`88.83% <94.73%> (ø)`	⬆️

Impacted Files	Coverage Δ
...c/Microsoft.ML.ImageAnalytics/ExtensionsCatalog.cs	`16.66% <ø> (+5.55%)`	⬆️
test/Microsoft.ML.Tests/ImagesTests.cs	`98.69% <94.73%> (-0.13%)`	⬇️
...ML.Transforms/Text/StopWordsRemovingTransformer.cs	`86.1% <0%> (-0.16%)`	⬇️
src/Microsoft.ML.Data/Transforms/Normalizer.cs	`86.03% <0%> (ø)`	⬆️
...icrosoft.ML.Functional.Tests/DataTransformation.cs	`100% <0%> (ø)`	⬆️
...s/Api/CookbookSamples/CookbookSamplesDynamicApi.cs	`93.49% <0%> (ø)`	⬆️
...s/Scenarios/Api/CookbookSamples/CookbookSamples.cs	`99.49% <0%> (ø)`	⬆️
test/Microsoft.ML.Functional.Tests/Training.cs	`100% <0%> (ø)`	⬆️
...est/Microsoft.ML.Tests/FeatureContributionTests.cs	`98.55% <0%> (ø)`	⬆️
...oft.ML.Experimental/TransformsCatalogExtensions.cs
... and 5 more

codecov · 2019-04-02T17:32:32Z

Codecov Report

Merging #3165 into master will increase coverage by 0.01%.
The diff coverage is 94.73%.

@@            Coverage Diff             @@
##           master    #3165      +/-   ##
==========================================
+ Coverage   72.53%   72.55%   +0.01%     
==========================================
  Files         808      807       -1     
  Lines      144775   144793      +18     
  Branches    16209    16211       +2     
==========================================
+ Hits       105012   105054      +42     
+ Misses      35348    35322      -26     
- Partials     4415     4417       +2

Flag	Coverage Δ
#Debug	`72.55% <94.73%> (+0.01%)`	⬆️
#production	`68.14% <ø> (+0.02%)`	⬆️
#test	`88.83% <94.73%> (ø)`	⬆️

Impacted Files	Coverage Δ
...c/Microsoft.ML.ImageAnalytics/ExtensionsCatalog.cs	`16.66% <ø> (+5.55%)`	⬆️
test/Microsoft.ML.Tests/ImagesTests.cs	`98.69% <94.73%> (-0.13%)`	⬇️
...ML.Transforms/Text/StopWordsRemovingTransformer.cs	`86.1% <0%> (-0.16%)`	⬇️
src/Microsoft.ML.Data/Transforms/Normalizer.cs	`86.03% <0%> (ø)`	⬆️
...icrosoft.ML.Functional.Tests/DataTransformation.cs	`100% <0%> (ø)`	⬆️
...s/Api/CookbookSamples/CookbookSamplesDynamicApi.cs	`93.49% <0%> (ø)`	⬆️
...s/Scenarios/Api/CookbookSamples/CookbookSamples.cs	`99.49% <0%> (ø)`	⬆️
test/Microsoft.ML.Functional.Tests/Training.cs	`100% <0%> (ø)`	⬆️
...est/Microsoft.ML.Tests/FeatureContributionTests.cs	`98.55% <0%> (ø)`	⬆️
...oft.ML.Experimental/TransformsCatalogExtensions.cs
... and 5 more

shmoradims · 2019-04-02T19:02:13Z

docs/samples/Microsoft.ML.Samples/Dynamic/Transforms/ImageAnalytics/ConvertToGrayScale.cs

+            foreach (var row in data.RowView)
+            {
+                foreach (var kvPair in row.Values)
+                {


{ [](start = 16, length = 1)

for 1-liner blocks, let's not use {} to keep the samples short. similar to foreach loop on line 58 #Resolved

actually these are 2-liners. may use Linq to shorten it.

In reply to: 271452481 [](ancestors = 271452481)

keeping 2 for loops to keep logic simple

In reply to: 271453121 [](ancestors = 271453121,271452481)

shmoradims · 2019-04-02T19:11:21Z

docs/samples/Microsoft.ML.Samples/Dynamic/Transforms/ImageAnalytics/ConvertToImage.cs

+        private static void PrintPreview(DataDebuggerPreview data)
+        {
+            foreach (var colInfo in data.ColumnView)
+                Console.Write("{0, -25}", colInfo.Column.Name);


"{0, -25}" [](start = 30, length = 10)

let's just define this as:
var format = "{0,-25}" (they usually don't use space after comma)

and use 'format' in the rest of the code

not getting you .. whats meant by using 'format' ? is the suggestion to use string.Format somehow ..

In reply to: 271455734 [](ancestors = 271455734)

used control+k+d to fix format (if thats what u meant)

In reply to: 271462084 [](ancestors = 271462084,271455734)

Ivan meant to define the "{0,-25}" as a variable, rather than repeating it, and reuse the variable below.

In reply to: 271465497 [](ancestors = 271465497,271462084,271455734)

shmoradims · 2019-04-02T19:11:49Z

docs/samples/Microsoft.ML.Samples/Dynamic/Transforms/ImageAnalytics/ConvertToImage.cs

+            var random = new Random(seed);
+
+            for (int i = 0; i < count; i++)
+            {


{ [](start = 12, length = 1)

omit #Resolved

sfilipi · 2019-04-03T06:36:25Z

docs/samples/Microsoft.ML.Samples/Dynamic/Transforms/ImageAnalytics/ConvertToImage.cs

+
+        private class DataPoint
+        {
+            [VectorType(inputSize)]


[VectorType(inputSize)] [](start = 11, length = 24)

is this annotation needed? #Resolved

yeah, this is indeed needed.

when fitting a model, we do some checks e.g. schema validation / size checks / whether the estimator works with fixed size or variable size etc

the VectorToImageConvertingEstimator works on fixed-size vector . without the annotation, the assumtion is that its a variable size vector , and the following check fails :

machinelearning/src/Microsoft.ML.ImageAnalytics/VectorToImageTransform.cs

Line 736 in a7bcdda

if (col.Kind != SchemaShape.Column.VectorKind.Vector || (col.ItemType != NumberDataViewType.Single && col.ItemType != NumberDataViewType.Double && col.ItemType != NumberDataViewType.Byte))

Unhandled Exception: System.ArgumentOutOfRangeException: Schema mismatch for input column 'Features': expected known-size vector of type Single, Double or Byte, got VarVector

In reply to: 271595813 [](ancestors = 271595813)

sfilipi

…3165) * updating image samples * fix review comments (print preview) * fix comments (minor nits) * fix review comments (whitespaces)

updating image samples

44870c5

abgoswam requested review from sfilipi and Ivanidzo4ka April 2, 2019 01:30

abgoswam changed the title ~~Image samples~~ Samples and unit test for image-related transform estimators Apr 2, 2019

abgoswam requested a review from shmoradims April 2, 2019 01:32

sfilipi reviewed Apr 2, 2019

View reviewed changes

fix review comments (print preview)

d5b4451

Ivanidzo4ka reviewed Apr 2, 2019

View reviewed changes

Ivanidzo4ka approved these changes Apr 2, 2019

View reviewed changes

fix comments (minor nits)

50c0f8d

sfilipi mentioned this pull request Apr 2, 2019

API reference - Samples for Transforms #1209

Closed

shmoradims reviewed Apr 2, 2019

View reviewed changes

fix review comments (whitespaces)

c8cbd15

sfilipi reviewed Apr 3, 2019

View reviewed changes

sfilipi approved these changes Apr 3, 2019

View reviewed changes

abgoswam merged commit ac53748 into dotnet:master Apr 3, 2019

abgoswam mentioned this pull request Apr 5, 2019

Fixing namespace for image samples #3208

Merged

abgoswam mentioned this pull request Apr 5, 2019

Cherry pick for samples {Image, Categorical, FeatureSelection} #3221

Closed

ghost locked as resolved and limited conversation to collaborators Mar 23, 2022


		// The transformedData IDataView contains the loaded and resized raw values

		// Preview 1 row of the transformedData.

		// Pixels : Dense vector of size 30000

		Console.WriteLine("--------------------------------------------------");


		var transformedData = pipeline.Fit(data).Transform(data);
		// The transformedData IDataView contains the loaded images now

Samples and unit test for image-related transform estimators #3165

Samples and unit test for image-related transform estimators #3165

Uh oh!

Conversation

abgoswam commented Apr 2, 2019

Uh oh!

sfilipi Apr 2, 2019 • edited by abgoswam Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sfilipi Apr 2, 2019 • edited by abgoswam Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sfilipi Apr 2, 2019 • edited by abgoswam Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sfilipi Apr 2, 2019 • edited by abgoswam Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sfilipi Apr 2, 2019 • edited by abgoswam Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sfilipi Apr 2, 2019 • edited by abgoswam Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sfilipi Apr 2, 2019 • edited by abgoswam Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sfilipi Apr 2, 2019 • edited by abgoswam Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sfilipi Apr 2, 2019 • edited by abgoswam Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Ivanidzo4ka Apr 2, 2019 • edited by abgoswam Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Ivanidzo4ka Apr 2, 2019 • edited by abgoswam Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Ivanidzo4ka left a comment

Choose a reason for hiding this comment

Uh oh!

codecov bot commented Apr 2, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

codecov bot commented Apr 2, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

shmoradims Apr 2, 2019 • edited by abgoswam Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

shmoradims Apr 2, 2019 • edited by sfilipi Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

sfilipi Apr 2, 2019 •

edited by abgoswam

Loading

sfilipi Apr 2, 2019 •

edited by abgoswam

Loading

sfilipi Apr 2, 2019 •

edited by abgoswam

Loading

sfilipi Apr 2, 2019 •

edited by abgoswam

Loading

sfilipi Apr 2, 2019 •

edited by abgoswam

Loading

sfilipi Apr 2, 2019 •

edited by abgoswam

Loading

sfilipi Apr 2, 2019 •

edited by abgoswam

Loading

sfilipi Apr 2, 2019 •

edited by abgoswam

Loading

sfilipi Apr 2, 2019 •

edited by abgoswam

Loading

Ivanidzo4ka Apr 2, 2019 •

edited by abgoswam

Loading

Ivanidzo4ka Apr 2, 2019 •

edited by abgoswam

Loading

codecov bot commented Apr 2, 2019 •

edited

Loading

codecov bot commented Apr 2, 2019 •

edited

Loading

shmoradims Apr 2, 2019 •

edited by abgoswam

Loading

shmoradims Apr 2, 2019 •

edited by sfilipi

Loading

shmoradims Apr 2, 2019 •

edited by abgoswam

Loading

sfilipi Apr 3, 2019 •

edited by abgoswam

Loading

abgoswam Apr 3, 2019 •

edited

Loading