ImageClassificationTrainer Class

Definition

Namespace:

Microsoft.ML.Vision

Assembly:

Microsoft.ML.Vision.dll

Package:

Microsoft.ML.Vision v3.0.1

The IEstimator<TTransformer> for training a Deep Neural Network(DNN) to classify images.

public sealed class ImageClassificationTrainer : Microsoft.ML.Trainers.TrainerEstimatorBase<Microsoft.ML.Data.MulticlassPredictionTransformer<Microsoft.ML.Vision.ImageClassificationModelParameters>,Microsoft.ML.Vision.ImageClassificationModelParameters>

type ImageClassificationTrainer = class
    inherit TrainerEstimatorBase<MulticlassPredictionTransformer<ImageClassificationModelParameters>, ImageClassificationModelParameters>

Public NotInheritable Class ImageClassificationTrainer
Inherits TrainerEstimatorBase(Of MulticlassPredictionTransformer(Of ImageClassificationModelParameters), ImageClassificationModelParameters)

Inheritance

Object

TrainerEstimatorBase<MulticlassPredictionTransformer<ImageClassificationModelParameters>,ImageClassificationModelParameters>

ImageClassificationTrainer

Remarks

To create this trainer, use ImageClassification.

Input and Output Columns

The input label column data must be key type and the feature column must be a variable-sized vector of Byte.

This trainer outputs the following columns:

Output Column Name	Column Type	Description
Score	Vector ofSingle	The scores of all classes.Higher value means higher probability to fall into the associated class. If the i-th element has the largest value, the predicted label index would be i.Note that i is zero-based index.
PredictedLabel	key type	The predicted label's index. If its value is i, the actual label would be the i-th category in the key-valued input label type.

Trainer Characteristics

Machine learning task	Multiclass classification
Is normalization required?	No
Is caching required?	No
Required NuGet in addition to Microsoft.ML	Microsoft.ML.Vision and SciSharp.TensorFlow.Redist / SciSharp.TensorFlow.Redist-Windows-GPU / SciSharp.TensorFlow.Redist-Linux-GPU
Exportable to ONNX	No

Using TensorFlow based APIs

In order to run any TensorFlow based ML.Net APIs you must first add a NuGet dependency on the TensorFlow redist library. There are currently two versions you can use. One which is compiled for GPU support, and one which has CPU support only.

CPU only

CPU based TensorFlow is currently supported on:

• Linux
• MacOS
• Windows

To get TensorFlow working on the CPU only all that is to take a NuGet dependency on SciSharp.TensorFlow.Redist v1.14.0

GPU support

GPU based TensorFlow is currently supported on:

• Windows
• Linux As of now TensorFlow does not support running on GPUs for MacOS, so we cannot support this currently.

Prerequisites

You must have at least one CUDA compatible GPU, for a list of compatible GPUs see Nvidia's Guide.

Install CUDA v10.1 and CUDNN v7.6.4.

Make sure you install CUDA v10.1, not any other newer version. After downloading CUDNN v7.6.4 .zip file and unpacking it, you need to do the following steps:

copy <CUDNN_zip_files_path>\cuda\bin\cudnn64_7.dll to <YOUR_DRIVE>\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\bin

For C/C++ development:

Copy <CUDNN_zip_files_path>\cuda\ include\cudnn.h to <YOUR_DRIVE>\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\include

Copy <CUDNN_zip_files_path>\cuda\lib\x64\cudnn.lib to <YOUR_DRIVE>\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\lib\x64

For further details in cuDNN you can follow the cuDNN Installation guide.

Usage

To use TensorFlow with GPU support take a NuGet dependency on the following package depending on your OS:

• Windows -> SciSharp.TensorFlow.Redist-Windows-GPU
• Linux -> SciSharp.TensorFlow.Redist-Linux-GPU

No code modification should be necessary to leverage the GPU for TensorFlow operations.

Troubleshooting

If you are not able to use your GPU after adding the GPU based TensorFlow NuGet, make sure that there is only a dependency on the GPU based version. If you have a dependency on both NuGets, the CPU based TensorFlow will run instead.

Training Algorithm Details

Trains a Deep Neural Network(DNN) by leveraging an existing pre-trained model such as Resnet50 for the purpose of classifying images. The technique was inspired from TensorFlow's retrain image classification tutorial

Fields

FeatureColumn	The feature column that the trainer expects. (Inherited from TrainerEstimatorBase<TTransformer,TModel>)
LabelColumn	The label column that the trainer expects. Can be null, which indicates that label is not used for training. (Inherited from TrainerEstimatorBase<TTransformer,TModel>)
WeightColumn	The weight column that the trainer expects. Can be null, which indicates that weight is not used for training. (Inherited from TrainerEstimatorBase<TTransformer,TModel>)

Properties

Info	Auxiliary information about the trainer in terms of its capabilities and requirements.

Methods

Fit(IDataView, IDataView)	Trains a ImageClassificationTrainer using both training and validation data, returns a ImageClassificationModelParameters.
Fit(IDataView)	Trains and returns a ITransformer. (Inherited from TrainerEstimatorBase<TTransformer,TModel>)
GetOutputSchema(SchemaShape)	(Inherited from TrainerEstimatorBase<TTransformer,TModel>)

Extension Methods

AppendCacheCheckpoint<TTrans>(IEstimator<TTrans>, IHostEnvironment)

Append a 'caching checkpoint' to the estimator chain. This will ensure that the downstream estimators will be trained against cached data. It is helpful to have a caching checkpoint before trainers that take multiple data passes.

WithOnFitDelegate<TTransformer>(IEstimator<TTransformer>, Action<TTransformer>)

Given an estimator, return a wrapping object that will call a delegate once Fit(IDataView) is called. It is often important for an estimator to return information about what was fit, which is why the Fit(IDataView) method returns a specifically typed object, rather than just a general ITransformer. However, at the same time, IEstimator<TTransformer> are often formed into pipelines with many objects, so we may need to build a chain of estimators via EstimatorChain<TLastTransformer> where the estimator for which we want to get the transformer is buried somewhere in this chain. For that scenario, we can through this method attach a delegate that will be called once fit is called.