ODModelBuilderTF 1.1.2

Suggested Alternatives

LG.ODModelBuilderTF 2.0.0

Additional Details

Packages starting without prefix LG. are no longer maintained. Please use the alternate package with the same functionalities.

dotnet add package ODModelBuilderTF --version 1.1.2
NuGet\Install-Package ODModelBuilderTF -Version 1.1.2
This command is intended to be used within the Package Manager Console in Visual Studio, as it uses the NuGet module's version of Install-Package.
<PackageReference Include="ODModelBuilderTF" Version="1.1.2" />
For projects that support PackageReference, copy this XML node into the project file to reference the package.
paket add ODModelBuilderTF --version 1.1.2
#r "nuget: ODModelBuilderTF, 1.1.2"
#r directive can be used in F# Interactive and Polyglot Notebooks. Copy this into the interactive tool or source code of the script to reference the package.
// Install ODModelBuilderTF as a Cake Addin
#addin nuget:?package=ODModelBuilderTF&version=1.1.2

// Install ODModelBuilderTF as a Cake Tool
#tool nuget:?package=ODModelBuilderTF&version=1.1.2


Object detection models builder with TensorFlow.

.NET (C#) object detection models builder with TensorFlow.

ODModelBuilderTF is a .NET library which allow to train object detection models directly in .NET environment, without the needing of Python code.<BR>

  • It's based on the TensorFlow framework.<BR>
  • It can be used with all .NET languages simply including the package on your project.<BR>
  • It's also event oriented, so it's possible to control the whole train process registering the events in the client application.<BR>
  • Tuning parameters are provided to manage the train process directly from the client without manually writing configuration files.<BR>
  • Can train and export TensorFlow's saved model format, frozen graph and ONNX models that can be afterwards consumed, for example, in the ML.NET framework.

Getting started with ODModelBuilderTF

Simply include the package (or the reference to the project if you include it in your solution) to used the library.<BR> At the first initialization it will install the train environment on your device downloading it from Internet (the process can take long time).<BR> If you prefer to have a ready to use environment (or your device is offline) you can include in your application the redistributables containing all the needed resources.

Sample apps

For a quick usage example you can take a look to the console application included in this repository, used to test the library.


LG.ODModelBuilderTF: the train library.<BR> LG.ODModelBuilderTF-Redist-Win: the redistributable with TensorFlow object detection packages.<BR> LG.ODModelBuilderTF-Redist-Win-TF: the redistributable with TensorFlow packages.

LG.ODModelBuilderTF-Redist-Win-CUDA10_1-TF: the redistributable with a special TensorFlow library for old GPUs which can work with CUDA 10.1 SM30. To use instead of the ODModelBuilderTF-Redist-Win-TF (CUDA 11) in old devices.<BR>

Simple installation with environment download at the first run:


Installation with full environment:

LG.ODModelBuilderTF, LG.ODModelBuilderTF-Redist-Win, LG.ODModelBuilderTF-Redist-Win-TF

Installation on old GPUs:

LG.ODModelBuilderTF, LG.ODModelBuilderTF-Redist-Win, LG.ODModelBuilderTF-Redist-Win-CUDA10_1-TF

Building ODModelBuilderTF

It's possible to build ODModelBuilderTF and the packages directly from the command line launching the build.cmd or opening the solution in Visual Studio 2019 or above.

Code examples

Here is a snippet code to show how to train a model.

using ODModelBuilderTF;
using System;
using System.Threading;
using System.Threading.Tasks;

namespace ODModelBuilderTF_Con
   class Program
      static async Task Main(string[] args)
         // Create the trainer
         var trainer = new Trainer(new Trainer.Options
            BatchSize = 16,
            CheckpointEvery = null,
            CheckpointForceThreashold = null,
            EvalImagesFolder = @"Data\Images\Eval",
            ExportFolder = @"Data\Export",
            ModelType = ModelTypes.SSD_MobileNet_V2_320x320,
            NumTrainSteps = 50000,
            OnnxModelFileName = "Model.onnx",
            TensorBoardPort = 6006,
            TrainFolder = @"Data\Train",
            TrainImagesFolder = @"Data\Images\Train",
            TrainRecordsFolder = @"Data\Records",
         // Step event
         var loss = default(double?);
         trainer.TrainStep += (sender, e) =>
            Console.WriteLine($"Step number:{e.StepNumber}\t\tstep time: {e.StepTime:N3} secs\t\ttotal loss:{e.TotalLoss:N3}");
            if (loss == null)
               loss = e.TotalLoss;
            else if (e.TotalLoss < loss) {
               loss = e.TotalLoss;
               e.CreateCheckpoint = true;
               Console.WriteLine($"{new string('=', 40)}> Create checkpoint with total loss {e.TotalLoss}");
         // Evaluation event
         var mAP = default(double?);
         trainer.TrainEvaluation += (sender, e) =>
            if (mAP == null)
               mAP = e.AP;
            else if (e.AP > mAP) {
               mAP = e.AP;
               e.Export = true;
            Console.WriteLine($"{new string('=', 40)}> Mean average precision {e.AP}");
            if (e.Export)
               Console.WriteLine($"{new string('=', 40)}> Exporting model...");
         // Export events
         trainer.ExportedSavedModel += (sender, e) =>
            Console.WriteLine($"{new string('=', 40)}> SavedModel exported");
         trainer.ExportedOnnx += (sender, e) =>
            Console.WriteLine($"{new string('=', 40)}> Onnx model exported");
         // Token for stop training
         var exitToken = new CancellationTokenSource();
         Console.CancelKeyPress += (sender, e) => exitToken.Cancel();
         // Wait the training
         await Task.Run(() => trainer.Train(exitToken.Token));

Colab notebook

It's possible to test / train directly on a google colab environment using the notebook ODModelBuilderTF.ipynb


ODModelBuilderTF is licensed under the MIT license and it is free to use commercially.

Product Compatible and additional computed target framework versions.
.NET net5.0 is compatible.  net5.0-windows was computed.  net6.0 was computed.  net6.0-android was computed.  net6.0-ios was computed.  net6.0-maccatalyst was computed.  net6.0-macos was computed.  net6.0-tvos was computed.  net6.0-windows was computed.  net7.0 was computed.  net7.0-android was computed.  net7.0-ios was computed.  net7.0-maccatalyst was computed.  net7.0-macos was computed.  net7.0-tvos was computed.  net7.0-windows was computed.  net8.0 was computed.  net8.0-android was computed.  net8.0-browser was computed.  net8.0-ios was computed.  net8.0-maccatalyst was computed.  net8.0-macos was computed.  net8.0-tvos was computed.  net8.0-windows was computed. 
Compatible target framework(s)
Included target framework(s) (in package)
Learn more about Target Frameworks and .NET Standard.

NuGet packages

This package is not used by any NuGet packages.

GitHub repositories

This package is not used by any popular GitHub repositories.

Version Downloads Last updated
1.1.2 524 8/14/2021
1.1.1 303 8/11/2021
1.1.0 295 8/9/2021