Multi platform (iOS, Androiod and UWP) mobile apps with Xamarin
Xamarin Certified Mobile Developer
Xamarin Certified Mobile Professional
El Románico Palentino Virtual Reality Tour (DEMO)
Unity Authorized Course
The Ultimate Guide to Game Development with Unity
3D virtual environments creator with XNA and HLSL
Natural User Interface applications with Kinect and Kinect v2 sensors
{videos}
Testing a deep neural network trained with NVIDIA Deep Learning GPU Training System (DIGITS) to recognize handwritten numbers
Augmented reality playing golf DEMO. Unity3D and Vuforia with image target
AR DEMO made with Vuforia Studio. An example simulating a sequence of instructions (invented) for repairing a kitchen using AR experience
AR DEMO made with Vuforia Studio. An example showing how to place a latch on a door in augmented reality
Augmented Reality DEMO. Testing ARCore SDK for Unity3D package
Data from this app, now with 3D virtual interface using MixedRealityToolkit-Unity, showing locations and monuments on Map, and showing monuments in 360 panoramic images (I will get the real images!, image showing is just an example..)
1. Intro -> Target:
Here, I'm going to resume the steps for making a simple test with one of the NVIDIA tools (DIGITS) for Deep Learning, train a neural network for image recognition, and develop a simple application as example for using the trained model and for showing the results returned.
The software requirements and required intallations are not indicated here, just the results, but links related to the material used, can be found in next steps.
2. NVIDIA DIGITS -> Deep Learning GPU Training System:
- The tool from NVIDIA used to train a deep neural network.
3. Train a model with MNIST handwritten digit dataset:
- The images dataset used for training the neural network is the MNIST database, a lot of images handwritten digits.
- Following these instrucctions, will help for using DIGITS to train the neural network with the MNIST dataset and how to export the trained model.
4. Using the trained model:
- Inside DIGITS source code, there are two python scripts that can help to use the trained model.
- Using Flask, a web microframework for building RESTful web API, receiving by POST an image with a digit picture to pass to the trained model.
5. The calculator application interface:
- Using WPF for making a Windows destop application, with two InkCavas controls to write on it a digit, capture the images of these controls, and send them to the runnig RESTful web API service.
-
The service returns the digit with the best confidence..
Obviously the sum operation is not the target :) is just an example of interface for showing the main purpose, check the trained neural network running..
Digits recognition (sum) with TensorFlow for a Xamarin App. A Binding Library built from TensorFlow Mobile, using TensorFlowInferenceInterface to recognize digits. Model trained with MNIST dataset.