Category: Artificial Intelligence
Emotiv MN8
Let us hope the SDK and raw EEG access is free with the new [Emotiv MN8] product.
First Deep Learning-Based 3D Simulation of the Universe
Tiny Machine Learning on the Edge with TensorFlow Lite Running on SAMD51
Neuroscience with Azure Machine Learning
TEDDY! – ARTIFICIAL INTELLIGENCE – (A.I)
Supercharge your Computer Vision models with the TensorFlow Object Detection API
Console App for Chrome Speech Proxy
I created a repo for [Console App for Chrome Speech Proxy] which is able to build for Mac and Windows. .Net Windows.Forms hasn’t been implemented for Mac yet…
Dr. Sbaitso in CodeName One
Shai Almog created a [demo app tutorial] that recreates the old “Dr. Sbaitso” app as a CodeName One project. “Dr. Sbaitso” was a text to speech demo app that shipped with Sound Blaster audio cards back in the 90s.
Source: [AI.java]
Source: [DrSbaitso.java]
Allen Cell Explorer
[Allen Cell Explorer] lets users browse detailed 3d scans of brain cells online. [nvidia]
Mastering StarCraft with AI
Visual Studio on Mac
The [Visual Studio Preview for Mac] is available.
Note: SCPlugin is not supported. [uninstall]
Visual Studio has detected that SCPlugin (scplugin.tigris.org) is installed. SCPlugin is a Subversion extension for Finder that is known to cause crashes in Visual Studio andother applications running on Mac OSX 10.9 (Mavericks) or upper. Please uninstall SCPlugin before proceeding.
Developer News:
[Cognitive Services – Enable Natural Language Interaction with LUIS]
Modeling with Speech Commands in Edit Mode in the Unity Editor
Documentation: [UnityWebGLSpeechDetection] [private]
Documentation: [UnityWebGLSpeechSynthesis] [private]
Unity: Speech Detection and Synthesis Together
To be able to call out “fire” and “stop”, I made some edits to the `F3DPlayerTurretController.cs` script.
using UnityEngine; using System.Collections; using UnityWebGLSpeechDetection; namespace Forge3D { public class F3DPlayerTurretController : MonoBehaviour { RaycastHit hitInfo; // Raycast structure public F3DTurret turret; bool isFiring; // Is turret currently in firing state public F3DFXController fxController; // reference to the proxy private ISpeechDetectionPlugin _mSpeechDetectionPlugin = null; enum FireState { IDLE, DETECTED_FIRE, FIRE_ONCE, FIRE_IDLE, DETECTED_STOP, STOP_ONCE } // detect the word once in all updates private static FireState _sFireState = FireState.IDLE; // make sure all turrets detect the async word in their update event private static bool _sReadyForLateUpdate = false; // init the speech proxy private IEnumerator Start() { // get the singleton instance _mSpeechDetectionPlugin = ProxySpeechDetectionPlugin.GetInstance(); // check the reference to the plugin if (null == _mSpeechDetectionPlugin) { Debug.LogError("Proxy Speech Detection Plugin is not set!"); yield break; } // wait for plugin to become available while (!_mSpeechDetectionPlugin.IsAvailable()) { yield return null; } // subscribe to events _mSpeechDetectionPlugin.AddListenerOnDetectionResult(HandleDetectionResult); // abort and clear existing words _mSpeechDetectionPlugin.Abort(); } // Handler for speech detection events void HandleDetectionResult(object sender, SpeechDetectionEventArgs args) { if (null == args.detectionResult) { return; } SpeechRecognitionResult[] results = args.detectionResult.results; if (null == results) { return; } bool doAbort = false; foreach (SpeechRecognitionResult result in results) { SpeechRecognitionAlternative[] alternatives = result.alternatives; if (null == alternatives) { continue; } foreach (SpeechRecognitionAlternative alternative in alternatives) { if (string.IsNullOrEmpty(alternative.transcript)) { continue; } string lower = alternative.transcript.ToLower(); Debug.LogFormat("Detected: {0}", lower); if (lower.Contains("fire")) { if (_sFireState == FireState.IDLE) { _sFireState = FireState.DETECTED_FIRE; } doAbort = true; } if (lower.Contains("stop")) { if (_sFireState == FireState.FIRE_IDLE) { _sFireState = FireState.DETECTED_STOP; } doAbort = true; } } } // abort detection on match for faster matching on words instead of complete sentences if (doAbort) { _mSpeechDetectionPlugin.Abort(); } } // make the async detected word, detectable at the start of all the update events void LateUpdate() { if (_sReadyForLateUpdate) { _sReadyForLateUpdate = false; switch (_sFireState) { case FireState.DETECTED_FIRE: _sFireState = FireState.FIRE_ONCE; break; case FireState.FIRE_ONCE: _sFireState = FireState.FIRE_IDLE; break; case FireState.DETECTED_STOP: _sFireState = FireState.STOP_ONCE; break; case FireState.STOP_ONCE: _sFireState = FireState.IDLE; break; } } } void Update() { CheckForTurn(); CheckForFire(); // After update, use one late update to detect the async word _sReadyForLateUpdate = true; } void CheckForFire() { // Fire turret //if (!isFiring && Input.GetKeyDown(KeyCode.Mouse0)) if (!isFiring && _sFireState == FireState.FIRE_ONCE) { isFiring = true; fxController.Fire(); } // Stop firing //if (isFiring && Input.GetKeyUp(KeyCode.Mouse0)) if (isFiring && _sFireState == FireState.STOP_ONCE) { isFiring = false; fxController.Stop(); } }
To be able to call out the names of weapons and to add speech, I made some edits to the `F3DFXController` script.
using System.Collections; using System; using UnityEngine; using UnityEngine.UI; using UnityWebGLSpeechDetection; using UnityWebGLSpeechSynthesis; namespace Forge3D { // Weapon types public enum F3DFXType { Vulcan, SoloGun, Sniper, ShotGun, Seeker, RailGun, PlasmaGun, PlasmaBeam, PlasmaBeamHeavy, LightningGun, FlameRed, LaserImpulse } public class F3DFXController : MonoBehaviour { /// <summary> /// Voices drop down /// </summary> public Dropdown _mDropdownVoices = null; /// <summary> /// Reference to the proxy /// </summary> private ISpeechDetectionPlugin _mSpeechDetectionPlugin = null; /// <summary> /// Reference to the proxy /// </summary> private ISpeechSynthesisPlugin _mSpeechSynthesisPlugin = null; /// <summary> /// Reference to the supported voices /// </summary> private VoiceResult _mVoiceResult = null; /// <summary> /// Reference to the utterance, voice, and text to speak /// </summary> private SpeechSynthesisUtterance _mSpeechSynthesisUtterance = null; /// <summary> /// Track when the utterance is created /// </summary> private bool _mUtteranceSet = false; /// <summary> /// Track when the voices are created /// </summary> private bool _mVoicesSet = false; enum WeaponState { IDLE, DETECTED_LEFT, LEFT_ONCE, DETECTED_RIGHT, RIGHT_ONCE } // detect the word once in all updates private static WeaponState _sWeaponState = WeaponState.IDLE; // make sure all turrets detect the async word in their update event private static bool _sReadyForLateUpdate = false; // Singleton instance public static F3DFXController instance; // init the speech proxy private IEnumerator Start() { // get the singleton instance _mSpeechDetectionPlugin = ProxySpeechDetectionPlugin.GetInstance(); // check the reference to the plugin if (null == _mSpeechDetectionPlugin) { Debug.LogError("Proxy Speech Detection Plugin is not set!"); yield break; } // wait for plugin to become available while (!_mSpeechDetectionPlugin.IsAvailable()) { yield return null; } _mSpeechSynthesisPlugin = ProxySpeechSynthesisPlugin.GetInstance(); if (null == _mSpeechSynthesisPlugin) { Debug.LogError("Proxy Speech Synthesis Plugin is not set!"); yield break; } // wait for proxy to become available while (!_mSpeechSynthesisPlugin.IsAvailable()) { yield return null; } // subscribe to events _mSpeechDetectionPlugin.AddListenerOnDetectionResult(HandleDetectionResult); // abort and clear existing words _mSpeechDetectionPlugin.Abort(); // Get voices from proxy GetVoices(); // Create an instance of SpeechSynthesisUtterance _mSpeechSynthesisPlugin.CreateSpeechSynthesisUtterance((utterance) => { //Debug.LogFormat("Utterance created: {0}", utterance._mReference); _mSpeechSynthesisUtterance = utterance; // The utterance is set _mUtteranceSet = true; // Set the default voice if ready SetIfReadyForDefaultVoice(); }); } /// <summary> /// Get voices from the proxy /// </summary> /// <returns></returns> private void GetVoices() { // get voices from the proxy _mSpeechSynthesisPlugin.GetVoices((voiceResult) => { _mVoiceResult = voiceResult; // prepare the voices drop down items SpeechSynthesisUtils.PopulateVoicesDropdown(_mDropdownVoices, _mVoiceResult); // The voices are set _mVoicesSet = true; // Set the default voice if ready SetIfReadyForDefaultVoice(); }); } /// <summary> /// Set the default voice if voices and utterance are ready /// </summary> private void SetIfReadyForDefaultVoice() { if (_mVoicesSet && _mUtteranceSet) { // set the default voice SpeechSynthesisUtils.SetDefaultVoice(_mDropdownVoices); // enable voices dropdown SpeechSynthesisUtils.SetInteractable(true, _mDropdownVoices); Voice voice = SpeechSynthesisUtils.GetVoice(_mVoiceResult, SpeechSynthesisUtils.GetDefaultVoice()); _mSpeechSynthesisPlugin.SetVoice(_mSpeechSynthesisUtterance, voice); // drop down reference must be set if (_mDropdownVoices) { // set up the drop down change listener _mDropdownVoices.onValueChanged.AddListener(delegate { // handle the voice change event, and set the voice on the utterance SpeechSynthesisUtils.HandleVoiceChanged(_mDropdownVoices, _mVoiceResult, _mSpeechSynthesisUtterance, _mSpeechSynthesisPlugin); }); } } } /// <summary> /// Speak the utterance /// </summary> private void Speak(string text) { if (!_mVoicesSet || !_mUtteranceSet) { // not ready return; } // Cancel if already speaking _mSpeechSynthesisPlugin.Cancel(); // Set the text that will be spoken _mSpeechSynthesisPlugin.SetText(_mSpeechSynthesisUtterance, text); // Use the plugin to speak the utterance _mSpeechSynthesisPlugin.Speak(_mSpeechSynthesisUtterance); } // Handler for speech detection events void HandleDetectionResult(object sender, SpeechDetectionEventArgs args) { if (null == args.detectionResult) { return; } SpeechRecognitionResult[] results = args.detectionResult.results; if (null == results) { return; } bool doAbort = false; foreach (SpeechRecognitionResult result in results) { SpeechRecognitionAlternative[] alternatives = result.alternatives; if (null == alternatives) { continue; } foreach (SpeechRecognitionAlternative alternative in alternatives) { if (string.IsNullOrEmpty(alternative.transcript)) { continue; } string lower = alternative.transcript.ToLower(); Debug.LogFormat("Detected: {0}", lower); if (lower.Contains("left")) { if (_sWeaponState == WeaponState.IDLE) { _sWeaponState = WeaponState.DETECTED_LEFT; } doAbort = true; break; } else if (lower.Contains("right")) { if (_sWeaponState == WeaponState.IDLE) { _sWeaponState = WeaponState.DETECTED_RIGHT; } doAbort = true; break; } else if (lower.Contains("lightning")) { if (DefaultFXType != F3DFXType.LightningGun) { DefaultFXType = F3DFXType.LightningGun; Speak(string.Format("{0} is active, sir", DefaultFXType)); } doAbort = true; break; } else if (lower.Contains("beam")) { if (DefaultFXType != F3DFXType.PlasmaBeam) { DefaultFXType = F3DFXType.PlasmaBeam; Speak(string.Format("{0} is active, sir", DefaultFXType)); } doAbort = true; break; } } } // abort detection on match for faster matching on words instead of complete sentences if (doAbort) { _mSpeechDetectionPlugin.Abort(); } } // make the async detected word, detectable at the start of all the update events void LateUpdate() { if (_sReadyForLateUpdate) { _sReadyForLateUpdate = false; switch (_sWeaponState) { case WeaponState.DETECTED_LEFT: _sWeaponState = WeaponState.LEFT_ONCE; break; case WeaponState.LEFT_ONCE: _sWeaponState = WeaponState.IDLE; break; case WeaponState.DETECTED_RIGHT: _sWeaponState = WeaponState.RIGHT_ONCE; break; case WeaponState.RIGHT_ONCE: _sWeaponState = WeaponState.IDLE; break; } } } void Update() { // Switch weapon types using keyboard keys //if (Input.GetKeyDown(KeyCode.RightArrow)) if (_sWeaponState == WeaponState.LEFT_ONCE) NextWeapon(); //else if (Input.GetKeyDown(KeyCode.LeftArrow)) if (_sWeaponState == WeaponState.RIGHT_ONCE) PrevWeapon(); // After update, use one late update to detect the async word _sReadyForLateUpdate = true; }
Unity: WebGL Speech Synthesis
My WebGL Speech Synthesis package has been accepted into the Unity Asset Store.
Unity: Switch Weapons With Speech
Unity: Drive Turrets With Speech
The [Sci-Fi Effects] assets comes with some great looking turrets and effects.
I used the [WebGL Speech Detection] package to add speech commands.
And to make speech work in the Unity editor, I added the [Chrome Speech Proxy].
To make Speech Detection work in the Turret example, I made some edits to the `F3DPlayerTurretController.cs` script.
// reference to the proxy private ProxySpeechDetectionPlugin _mProxySpeechDetectionPlugin = null; enum FireState { IDLE, DETECTED_FIRE, FIRE_ONCE, FIRE_IDLE, DETECTED_STOP, STOP_ONCE } // detect the word once in all updates private static FireState _sFireState = FireState.IDLE; // make sure all turrets detect the async word in their update event private static bool _sReadyForLateUpdate = false; // init the speech proxy private IEnumerator Start() { while (null == WebGLSpeechDetectionPlugin.GetInstance() || null == ProxySpeechDetectionPlugin.GetInstance() || !ProxySpeechDetectionPlugin.GetInstance().IsAvailable()) { yield return null; } // reference to the plugin WebGLSpeechDetectionPlugin plugin = WebGLSpeechDetectionPlugin.GetInstance(); // subscribe to events plugin.OnDetectionResult += HandleDetectionResult; // reference to the proxy _mProxySpeechDetectionPlugin = ProxySpeechDetectionPlugin.GetInstance(); // abort and clear existing words _mProxySpeechDetectionPlugin.Abort(); } // Handler for speech detection events void HandleDetectionResult(object sender, WebGLSpeechDetectionPlugin.SpeechDetectionEventArgs args) { if (null == args.detectionResult) { return; } WebGLSpeechDetectionPlugin.SpeechRecognitionResult[] results = args.detectionResult.results; if (null == results) { return; } bool doAbort = false; foreach (WebGLSpeechDetectionPlugin.SpeechRecognitionResult result in results) { WebGLSpeechDetectionPlugin.SpeechRecognitionAlternative[] alternatives = result.alternatives; if (null == alternatives) { continue; } foreach (WebGLSpeechDetectionPlugin.SpeechRecognitionAlternative alternative in alternatives) { if (string.IsNullOrEmpty(alternative.transcript)) { continue; } string lower = alternative.transcript.ToLower(); Debug.LogFormat("Detected: {0}", lower); if (lower.Contains("fire")) { if (_sFireState == FireState.IDLE) { _sFireState = FireState.DETECTED_FIRE; } doAbort = true; } if (lower.Contains("stop")) { if (_sFireState == FireState.FIRE_IDLE) { _sFireState = FireState.DETECTED_STOP; } doAbort = true; } } } // abort detection on match for faster matching on words instead of complete sentences if (doAbort) { _mProxySpeechDetectionPlugin.Abort(); } } // make the async detected word, detectable at the start of all the update events void LateUpdate() { if (_sReadyForLateUpdate) { _sReadyForLateUpdate = false; switch (_sFireState) { case FireState.DETECTED_FIRE: _sFireState = FireState.FIRE_ONCE; break; case FireState.FIRE_ONCE: _sFireState = FireState.FIRE_IDLE; break; case FireState.DETECTED_STOP: _sFireState = FireState.STOP_ONCE; break; case FireState.STOP_ONCE: _sFireState = FireState.IDLE; break; } } } void Update() { CheckForTurn(); CheckForFire(); // After update, use one late update to detect the async word _sReadyForLateUpdate = true; } void CheckForFire() { // Fire turret //if (!isFiring && Input.GetKeyDown(KeyCode.Mouse0)) if (!isFiring && _sFireState == FireState.FIRE_ONCE) { isFiring = true; fxController.Fire(); } // Stop firing //if (isFiring && Input.GetKeyUp(KeyCode.Mouse0)) if (isFiring && _sFireState == FireState.STOP_ONCE) { isFiring = false; fxController.Stop(); } }
To be able to call out the names of weapons, I made some edits to the `F3DFXController` script.
// reference to the proxy private ProxySpeechDetectionPlugin _mProxySpeechDetectionPlugin = null; enum WeaponState { IDLE, DETECTED_LEFT, LEFT_ONCE, DETECTED_RIGHT, RIGHT_ONCE } // detect the word once in all updates private static WeaponState _sWeaponState = WeaponState.IDLE; // make sure all turrets detect the async word in their update event private static bool _sReadyForLateUpdate = false; // Singleton instance public static F3DFXController instance; // init the speech proxy private IEnumerator Start() { while (null == WebGLSpeechDetectionPlugin.GetInstance() || null == ProxySpeechDetectionPlugin.GetInstance() || !ProxySpeechDetectionPlugin.GetInstance().IsAvailable()) { yield return null; } // reference to the plugin WebGLSpeechDetectionPlugin plugin = WebGLSpeechDetectionPlugin.GetInstance(); // subscribe to events plugin.OnDetectionResult += HandleDetectionResult; // reference to the proxy _mProxySpeechDetectionPlugin = ProxySpeechDetectionPlugin.GetInstance(); // abort and clear existing words _mProxySpeechDetectionPlugin.Abort(); } // Handler for speech detection events void HandleDetectionResult(object sender, WebGLSpeechDetectionPlugin.SpeechDetectionEventArgs args) { if (null == args.detectionResult) { return; } WebGLSpeechDetectionPlugin.SpeechRecognitionResult[] results = args.detectionResult.results; if (null == results) { return; } bool doAbort = false; foreach (WebGLSpeechDetectionPlugin.SpeechRecognitionResult result in results) { WebGLSpeechDetectionPlugin.SpeechRecognitionAlternative[] alternatives = result.alternatives; if (null == alternatives) { continue; } foreach (WebGLSpeechDetectionPlugin.SpeechRecognitionAlternative alternative in alternatives) { if (string.IsNullOrEmpty(alternative.transcript)) { continue; } string lower = alternative.transcript.ToLower(); Debug.LogFormat("Detected: {0}", lower); if (lower.Contains("left")) { if (_sWeaponState == WeaponState.IDLE) { _sWeaponState = WeaponState.DETECTED_LEFT; } doAbort = true; } else if (lower.Contains("right")) { if (_sWeaponState == WeaponState.IDLE) { _sWeaponState = WeaponState.DETECTED_RIGHT; } doAbort = true; } else if (lower.Contains("lightning")) { DefaultFXType = F3DFXType.LightningGun; doAbort = true; } else if (lower.Contains("beam")) { DefaultFXType = F3DFXType.PlasmaBeam; doAbort = true; } } } // abort detection on match for faster matching on words instead of complete sentences if (doAbort) { _mProxySpeechDetectionPlugin.Abort(); } } // make the async detected word, detectable at the start of all the update events void LateUpdate() { if (_sReadyForLateUpdate) { _sReadyForLateUpdate = false; switch (_sWeaponState) { case WeaponState.DETECTED_LEFT: _sWeaponState = WeaponState.LEFT_ONCE; break; case WeaponState.LEFT_ONCE: _sWeaponState = WeaponState.IDLE; break; case WeaponState.DETECTED_RIGHT: _sWeaponState = WeaponState.RIGHT_ONCE; break; case WeaponState.RIGHT_ONCE: _sWeaponState = WeaponState.IDLE; break; } } } void Update() { // Switch weapon types using keyboard keys //if (Input.GetKeyDown(KeyCode.RightArrow)) if (_sWeaponState == WeaponState.LEFT_ONCE) NextWeapon(); //else if (Input.GetKeyDown(KeyCode.LeftArrow)) if (_sWeaponState == WeaponState.RIGHT_ONCE) PrevWeapon(); // After update, use one late update to detect the async word _sReadyForLateUpdate = true; }
Chrome Speech Proxy
The [Chrome Speech Proxy] uses the Chrome Browser for the Speech API to do real-time speech detection without any quotas. This makes Speech Detection available on Windows and in the Unity editor.
Emotiv: Community SDK
Emotiv has a free [Community SDK] for interacting with the Insight and Epoc headsets.
The developer community hangs out on the [forums] and in the [G+ Community].
Unreal: Speech Plugins
I created repositories to hold some new Unreal projects.
Documentation: [UnrealHTML5SpeechDetection] [private repo]
Research:
Default Local Build: [localhost:8000]
[Unreal: HTML5 – Getting Started]
[How to reduce HTML5 package size?]
Files Required for Final Deployment ----------------------------------- *.js.gz - compressed JavaScript files. *.data - compressed game content. *.mem - compressed memory initialization file. *.html - uncompressed landing page. *.symbols - uncompressed symbols, if necessary.
[HTML5: Call C++ UFUNCTION from Webpage via JavaScript]
[HTML5Platform.Automation.cs] includes [GameX.html.template]
[Connecting C++ and HTML5] [info]
Video: [Getting started with Emscripten – Transpiling C / C++ to JavaScript / HTML5]
Issues:
Pull Request: [Add support for including project and plugin JS]
Microsoft Debuts Customizable Speech-To-Text Tech, Releases Some Cognitive Services Tools To Developers
Unity – Speech Plugins
I created a public repositories to hold the online documentation for Unity speech related packages.
[Demo 01 Unity Speech Dictation]
[Demo 02 Unity Speech Commands]
[Demo 01 Unity Speech Synthesis]
Documentation: [UnityWebGLSpeechDetection] [private]
Documentation: [UnityWebGLSpeechSynthesis] [private]
Documentation and Source: [UnityWebGLMicrophone]
Documentation: [WebGL: Interacting with browser scripting]
The [Chrome Speech Demo] supports multiple languages.
[UnityWebGLDropdown Test] [repo] [issue]
[LOW LEVEL PLUGINS IN UNITY WEBGL]
Jasper Voice Control
[Jasper] is an open source platform for developing always-on, voice-controlled applications.