Easy to use
DLHUB does not require programming skills to design a deep learning neural network model for your application. Friendly graphical user interface with drag-and-drop features that helps to construct deep learning model with few simple clicks.
DLHUB supports advanced training algorithms, including AdaDelta, AdaGrad, FSAdaGrad, Adam, RMSProp, MomentumSGD, and SGD with mini-batch support. DLHUB also supports GPU to speed-up training process.
DLHUB supports evaluating trained deep learning model with new dataset to give a designer an instant feedback of how good the trained model can cope with unknown data. As a result, neccesary tweak can be made to improve model stability.
DLHUB supports different types of nodes, including Conv2D, Pool2D, Dense, Flatten, Drop, ResNet, Activation, TLModel (transfer learning), and recurrent nodes to allow a designer to customise a deep neural network structure.
Easy to deploy
Trained neural network model can be easily deployed in real-time applications developed various programming languages such as LabVIEW, LabVIEW NXG, LabWindow CVI, C.C++, and C# by using supported APIs provided by ANS CENTER.
DLHUB supports different dataset formats to simplify data processing task. DLHUB supports template scripts that helps a designer choose an appropriate deep learning architecture for a given application and dataset type.
DLHUB supports different type of dataset, including text file, image files, and image folders. When dataset is loaded into DLHUB, data normalisation procedure is automatically applied.
Configure a deep learning neural network structure
DLHUB supports various deep learning node types that helps constructing any deep learning neural network model with ease. A designer needs to select node type (1), modify its parameters (2), re-arrange deep learning architecture (3), and verify it this deep learning model is valid or not (4). During designing process, designer can save and/or load this model configuration into cfg file. DLHUB also provides model templates with different deep learning architectures that helps a designer speed-up designing task.
Train a deep learning neural network
DLHUB supports various training algorithms to allow multiple design options for a certain problems. Designer needs to select appropriate training algorithm with its paramters (1), and stopping criteria (2) to decide when to stop before start training process (3). Graphical user interface allows designer access visual feedback during training process. Designer can stop training at any time to tweak training parameters. GPU is automatically dectected and enabled to speed-up training process.
Evaluate a deep learning neural network
DLHUB supports evaluating a trained deep learning model with new dataset. That helps designer to validate the deep learning model in different test scenarios before deciding to use the trained model in an actual production enviroment/application.
Test trained deep learning neural network model for deployment.
Although evaluation helps designer selecting correct trained model architecture, it is useful to test the trained model again with completely unknown test data to see how the trained model behaves before exporting it for deployment.
Support transfer learning
DLHUB supports transfer learning that can re-use proven-to-work trained deep learning models.
DLHUB supports loading different test dataset format. For image data, folder option is supported.
Learn DLHUB via case studies
ANS Center provides elegent solutions to simplify machine learning and deep learning design and deployment.
We support mutiple hardware platforms and programming languages, including LabVIEW, LabVIEW NXG, LabWindow CVI, C/C++/C#, and Arduino.