3. Luis Beltrán
• Investigador en Tomás Bata University in Zlín,
República Checa
• Docente en Tecnológico Nacional de México en
Celaya
• Xamarin, Azure, Inteligencia Artificial
@darkicebeam
luis@luisbeltran.mx
4. Computer Vision
Face API
Custom Vision
Text-to-Speech
Speech-to-Text
Speech Translator
Speaker Recognition
Language Understanding
Text Translator
Text Analytics
QnA Maker
Anomaly Detector
Content Moderator
Personalizer
microsoft.com/cognitive
Visión Voz Lenguaje Decisión
Poniendo la Inteligencia Artificial al alcance de
los desarrolladores sin requerir experiencia previa
en machine learning
Azure Cognitive Services
5. Azure Applied AI Services
• Son servicios de alto nivel centrados en
ayudar a los desarrolladores a liberar en
poco tiempo el valor de sus datos en
escenarios empresariales aplicando IA.
• Están basados en las APIs de los Servicios
Cognitivos de Azure.
7. Imagina si pudieras…
• Obtener automáticamente metadatos (texto hablado, texto
escrito, caras, lugares, objetos, etc.) de cualquier video.
• Buscar información específica en audios, por ejemplo,
detectar clientes frustrados o contar el tiempo de espera en
un centro de llamadas.
• Cargar y editar videos de manera colaborativa en tiempo
real, buscando momentos clave y acelerando el tiempo
para publicar contenido nuevo.
• Permitir que sus espectadores busquen contenido
multimedia por nombre, palabras clave, ubicación física,
objetos.
• Distribuir clips de contenido de video en vivo sin tiempos
prolongados de descarga o codificación.
• Crear resúmenes automatizados o resaltar secciones de un
video en función de escenas, personas específicas y
movimiento.
• Traducir el contenido de un video a otros idiomas para
alcanzar una audiencia global y diversa.
8. La IA aplicada en multimedia es la clave para
resolver estos desafíos
• Mejora la visibilidad de contenido
• Incrementa el valor del contenido
• Personaliza la experiencia de
visualización
• Descubre información sobre
contenido oculto
• Reduce el trabajo manual
• Aumenta los ingresos
• Promueve la publicidad dirigida
• Genera modelado predictivo y
recomendaciones
9. • Azure Video Analyzer es un servicio que extrae detalles
valiosos de un archivo multimedia.
• Utiliza modelos de machine learning que pueden ser
personalizados y reentrenados.
• Los detalles del video incluyen detección de rostros,
reconocimiento de texto, etiquetado de objetos,
segmentación de escenas, etc.
• Se extraen detalles adicionales del audio, como la
transcripción y la detección de emociones.
• Puedes usar estos resultados para mejorar la búsqueda,
extraer clips, crear miniaturas y más, mejorando así la
participación del usuario.
• Está disponible como servicio y como plataforma.
Azure Video Analyzer Azure Video Analyzer
Analiza los canales visuales y de audio en un video,
indexando su contenido.
10. Suscríbete al API
• Para usar el servicio de Azure Video
Analyzer se requiere una suscripción.
• Puedes crear una cuenta de tipo trial
(gratuita), la cual incluye:
• Hasta 600 minutos de indexación
gratuita mediante el Portal de
Azure Video Analyzer
• Hasta 2400 minutos gratuitos
utilizando el API.
• Una suscripción disponible para
ser utilizada en el servicio.
videoindexer.ai
11. Azure Video Analyzer for Media Portal
• Sube videos
• Desde una dirección URL
• Desde archivos
• Los formatos multimedia
más comunes están
soportados:
• MOV
• WMV
• MPG
• AVI
• MP4
• ...
• Indexa características
30. Call to Action
• Video Analyzer for Media Portal
https://www.videoindexer.ai/
• Video Analyzer for Media Developer Portal
https://api-portal.videoindexer.ai/
• Microsoft Learn: Create computer vision solutions with Azure
Cognitive Services – Analyze Video
https://docs.microsoft.com/en-us/learn/modules/analyze-video/
32. ¡Gracias por tu atención!
Luis Beltrán
Tomás Bata University in Zlín
Tecnológico Nacional de México en Celaya
luis@luisbeltran.mx luisbeltran.mx @darkicebeam
About Me:
https://about.me/luis-beltran
Notas del editor
Azure Cognitive Services is a suite of services and APIs backed by machine learning that enables developers to incorporate intelligent features such as facial recognition in photos and videos, sentiment analysis in text, and language understanding into their applications.
Video Indexer is one of the newest members of the Cognitive Services family. Its purpose is to transform raw video content into content that is searchable, discoverable, and more engaging to the user. Want to generate a video transcript, index words spoken in the video or written on a whiteboard, or create a list of keywords from topics discussed in the video? Video Indexer can do all this and more. It can even find individuals in the video, and sometimes tell who they are.
Video Indexer is both a service and an API. The service is accessed through a Web portal. It allows you to upload videos and examine the information generated from them. The Video Indexer API is a REST API that does everything the portal does, and also allows you to access the information that is generated when videos are indexed.
Open the Video Indexer portal in your browser and select Get Started. Then select Sign in with your work account (AAD) to sign in with a work or school account, or Sign in with your personal Microsoft account to use a personal Microsoft account. Answer Yes if prompted to let this app access your info.
Select the Upload button. Then select enter a file url to upload a video from a URL.
Paste the following URL into the URL field, and enter "Overview of the Microsoft AI School" as the video name. Then select Upload to begin the upload.
https://topcs.blob.core.windows.net/public/Machine-Learning-in-IoT-solutions_high.mp4
Machine Learning in IoT Solutions
When indexing is complete, you will receive an e-mail notification for each video with a link to the video and a short description of what was found in it, such as people, topics, and keywords. Wait for all three videos to finish indexing, and then proceed to the next exercise.
Video Indexer provides regarding each video that it indexes, and this information is available not only in the portal, but through the Video Indexer API.
In Video Indexer, insights are aggregated views of the knowledge extracted from a video, such as faces, keywords, and sentiment. For example, you can see the faces of people appearing in the video, as well as time ranges and percentages for each face shown. Video Indexer cross-references the faces that it finds against a database of thousands of famous people and automatically identifies them. You can see for yourself by opening the "Microsoft in Education" video in the portal. Microsoft CEO Satya Nadella appears in that video, and Video Indexer recognizes him.
Video Indexer automatically generates video transcripts based on its built-in speech and speaker recognition services. It even provides facilities for editing the information that it generated so you can correct errors in transcripts, put names to faces that weren't recognized, and more.
The "Insights" tab shows people featured in the video, keywords generated from the video, topics identified in the video, brands featured in the video, and even emotions found in the video. You can select Play next for any of these items and cycle through the corresponding points in the video.
In this example, Video indexer found two people in the video. It was unable to identify them because they don't appear in its database of famous people. However, you can lend a helping hand by identifying them yourself. Enable editing by selecting the Edit icon in the upper-right corner. Then select the pencil icon next to "Unknown #1" and enter "Sonya Koptyev" as the person's name. Finish up by pressing Enter to save the change.
Repeat this step for the "Unknown #2" in the video. This person's name is "Seth Juarez".
Want to see a full transcript of the video? Select Timeline at the top of the page. Video Indexer uses a deep neural network (DNN) to aid in converting speech to text, but such conversions are rarely perfect. Here, too, you can help out by editing words and phrases that weren't converted properly. To demonstrate, make sure you're still in editing mode and change "High Amsonia captive." to "Hi, I'm Sonya Koptyev.”
Select Insights, and then search for the word "intelligence." This time, the results are conceptual topics that include the search term.
Video Indexer has the ability to translate transcripts into a variety of languages, including German, Dutch, Spanish, French, Czech, Korean, and Japanese. To demonstrate, select Timeline again, select the world icon, and select a language other than English from the drop-down list.
The "Insights" tab shows people featured in the video, keywords generated from the video, topics identified in the video, brands featured in the video, and even emotions found in the video. You can select Play next for any of these items and cycle through the corresponding points in the video.
In this example, Video indexer found two people in the video. It was unable to identify them because they don't appear in its database of famous people. However, you can lend a helping hand by identifying them yourself. Enable editing by selecting the Edit icon in the upper-right corner. Then select the pencil icon next to "Unknown #1" and enter "Sonya Koptyev" as the person's name. Finish up by pressing Enter to save the change.
Repeat this step for the "Unknown #2" in the video. This person's name is "Seth Juarez".
Want to see a full transcript of the video? Select Timeline at the top of the page. Video Indexer uses a deep neural network (DNN) to aid in converting speech to text, but such conversions are rarely perfect. Here, too, you can help out by editing words and phrases that weren't converted properly. To demonstrate, make sure you're still in editing mode and change "High Amsonia captive." to "Hi, I'm Sonya Koptyev.”
Select Insights, and then search for the word "intelligence." This time, the results are conceptual topics that include the search term.
Video Indexer has the ability to translate transcripts into a variety of languages, including German, Dutch, Spanish, French, Czech, Korean, and Japanese. To demonstrate, select Timeline again, select the world icon, and select a language other than English from the drop-down list.
Want to see a full transcript of the video? Select Timeline at the top of the page. Video Indexer uses a deep neural network (DNN) to aid in converting speech to text, but such conversions are rarely perfect. Here, too, you can help out by editing words and phrases that weren't converted properly. To demonstrate, make sure you're still in editing mode and change "High Amsonia captive." to "Hi, I'm Sonya Koptyev.”
Select Insights, and then search for the word "intelligence." This time, the results are conceptual topics that include the search term.
Video Indexer has the ability to translate transcripts into a variety of languages, including German, Dutch, Spanish, French, Czech, Korean, and Japanese. To demonstrate, select Timeline again, select the world icon, and select a language other than English from the drop-down list.
Once a video is indexed, you can search its contents. Type "suggestion" into the search box at the top of the page and press Enter. Confirm that the search results include four instances in which the word "suggestion" was found in the video.
Select Insights, and then search for the word "intelligence." This time, the results are conceptual topics that include the search term.
The Video Indexer portal provides a window into the videos that you index and lets you see a wealth of information extracted from them. But the real power of Video Indexer lies in the Video Indexer API, which lets you submit videos for indexing programmatically and access the results using a REST API. In Exercise 4, you will build an app that uses this API to expose content in the videos you indexed in Exercise 1. But to call the API, you must first subscribe to it and obtain an API key that is transmitted in each request. In this exercise, you will create a Video Indexer API subscription and retrieve the API key created for it.
Open the Video Indexer API portal in your browser and select SIGN IN in the top-right corner. Sign in with your Microsoft account — the same one you used to sign in to the Video Indexer portal. Answer Yes if prompted to let this app access your info.
Select Products, and then select Authorization.
Select the Subscribe button. The subscription will be created.
Select Products again, and then select Authorization. You will now see a list of the subscriptions you have, so select the Product Authorization Subscription.
Select the Subscribe button. The subscription will be created.
Select Products again, and then select Authorization. You will now see a list of the subscriptions you have, so select the Product Authorization Subscription.
Select the Show button next to the Primary Key. Copy the API key to the clipboard, and then select Hide to hide it again.
Now that you have an API key, you can write apps that call the Video Indexer API. The API key travels in an HTTP header in each request. Without a valid API key, the Video Indexer API fails requests placed to it. It is the API's way of ensuring that the caller is authorized.
Open the Video Indexer API portal in your browser and select SIGN IN in the top-right corner. Sign in with your Microsoft account — the same one you used to sign in to the Video Indexer portal. Answer Yes if prompted to let this app access your info.
Select Products, and then select Authorization.
Select the Subscribe button. The subscription will be created.
Select Products again, and then select Authorization. You will now see a list of the subscriptions you have, so select the Product Authorization Subscription.
Select the Subscribe button. The subscription will be created.
Select Products again, and then select Authorization. You will now see a list of the subscriptions you have, so select the Product Authorization Subscription.
Select the Show button next to the Primary Key. Copy the API key to the clipboard, and then select Hide to hide it again.
Now that you have an API key, you can write apps that call the Video Indexer API. The API key travels in an HTTP header in each request. Without a valid API key, the Video Indexer API fails requests placed to it. It is the API's way of ensuring that the caller is authorized.
There is no cost associated with the AI part because it doesn't require an Azure subscription.
The Video Indexer API is a rich one that includes methods for uploading videos for indexing, searching indexed videos, retrieving and modifying transcripts, monitoring the processing state as a video is being indexed, and more. One of the more powerful methods is Get Video Index, which returns the indexed content of a video containing the same kind of detailed information found in the Video Indexer portal after a video is indexed.
Video Explorer uses the Search Videos method, which is just one of more than 30 methods featured in the Video Indexer API. You could leverage additional APIs to make Video Explorer richer and more interactive. For example, you could allow users to search for people that appear in a video, or search specifically for text that is extracted via OCR. You could even use the Get Video Player Widget URL method to embed a video player in the app. Feel free to use these APIs to expand Video Explorer and customize it to fit your needs, and learn more about Video Indexer in the process.