Introduction to Artificial Intelligence on Azure


Artificial intelligence is a technology that equips machines to think and emulate intelligently like human beings. It is a software that helps the machines to imitate human behaviour and capabilities. The machines are fed some data points related to the scenario and this then allows them to predict, analyze and arrive at conclusions similar to humans. This intelligence is used to detect abnormal events and occurrences. Machines are able to detect and interpret visual inputs like recognizing objects and people. Visual interpretation and intelligence are key aspects of human intelligence and with some data inputs and algorithms, we can bring on the same intelligence to machines as well. Artificial intelligence also helps machines interpret human spoken and written languages. With appropriate training, the machines can converse and communicate effectively. It can be used to engage in dialogues and conversations and can be made to do routine tasks of customer support and customer service. Machine learning provides the ability for the machines to train and learn from experience to exhibit this artificial intelligence. Now we are going to see different services and artificial intelligence workloads within Azure.


Machine learning


Machine learning is part of artificial intelligence wherein it creates models and brings about predictions based on the relationship between the data that it has been provided. The model analyses the data and builds a relationship between different data points and learns from these points. Once new data is provided, the model then makes predictions based on the learning. For example, assume that we are in a garden with beautiful flowers of different shapes, sizes and colours. We manage to learn the name of every flower with the help of a botanist. We then feed this data which contains the images of the flowers in different sizes and colours aligned with their names and build a model. The resulting machine learning model will then be seeing a flower and then make a prediction based on the data points that have been fed and trained. This is how different AI services function using machine learning models and create predictions.

Anomaly detection

Anomaly detection is a very crucial aspect of artificial intelligence. It can be powerful and actually save lives. Assume that you have a detection system in your car to track your tyre pressure, your engine temperature or brake health. This system will be tracking your data points every single day. One day when things are matching up to the measured data points like when your brakes are not working well, or if your type pressure numbers are not matching, then the system will alert you or your driver about the anomaly. This can be really useful and can prevent accidents and divert attention to the anomaly before it causes any catastrophe.

Computer Vision


The purpose behind AI is to give the machine the power to emulate the human experience. Computer vision gives the machines a way to view the world as we see it. The input to such a system is thousands of images as the eyes daily see. These images are made up of millions of pixels and each pixel intensity ranges from 0 to 255. Pixel and intensity changes are the way the machines learn to analyze the images. helpThe images are further broken down into its RGB channels and then the computer makes detections and classifications based on the image. It may look very complicated but the way the machine understands it is different, and it is the basis in which it makes its predictions. Computer vision can be used for image classification to classify a particular type of vehicle as a taxi based on certain distinctive features and points that are provided to the machine.

Objection detection using computer vision is interesting. Imagine a busy scenario as a street or a road, where we have multiple objects like a pedestrian, cyclist, buses, cars, taxi, the object detection system finds all the objects correctly even when only part of the object is observed. For example, in the following picture, we find on the top left corner a part of the bus that is shown and the detection algorithm perfectly identifies it as a bus.

Semantic Segmentation is an advanced system wherein the algorithm does not only identify the object through a bounding box, but it also identifies the individual pixels and masks the object correctly. This can be used to differentiate between models of vehicles and their make. It can also be used to detect different objects in a photograph and colour code them for our use. Here you can see in the image that a different colour code is given for different objects in the image

Image Analysis is a very intelligent detection or analysing system. In the following example image, the algorithm not only detects a man, dog and other objects, but it also makes a correct analysis of the man with a dog on the street. The specifics of the image is correctly detected and pointed out. The extent to which the details are analysed is amazing.

Using computer vision, we can detect faces and even recognize them. We can predict their age, gender, their gestures and emotions including their distinct features like moustache and the colour of their lipstick. We have a free API demo for face detection on the azure website which can be used in our own environment. With image analysis and face detection, a lot can be achieved. There is a case study where visually impaired use computer vision trained hololens to detect faces and familiar people. As they are scanning the room, they can know who is coming or approaching the person based on the image analysis. It is very powerful for those who are visually impaired and help them understand their surroundings better.

With Optical Character recognition using computer vision, images can be analysed and text or signs on the same can be detected and understood. When recordings are grainy and blurry, using computer vision we can analyse these photos and images and detect the signs, license plates and any text on the image.

Security and safety issues have also been addressed by the use of computer vision. For example travel by Uber and cabs have been rendered safe using AI. The drivers or the passengers can be asked to take selfies and their faces can be matched to the information provided during booking to ensure a safe ride for both the driver and the passenger. Next, we move on to other services supported by Azure in artificial intelligence.

Natural Language Processing


We now discuss some of the Cognitive services provided by Azure. Text analytics where it recognizes a language that is spoken based on typing or scanning documents. Next is speech recognition, where we speak to a device and give it instructions and the device is able to interpret the language spoken and it translates it to the action that needs to be performed. Real-time translation of speech to text is a very exciting prospect of AI. It allows people of different language and origin to converse comfortably using speech to text cognitive service provided by the PowerPoint translator. One of the ways cognitive services can support translation services is by translating machine manuals into different languages for companies that have a global presence. Language understanding is something we all use on a daily basis, where we use smart devices in the house. For example, we can turn off or on the light or AC or thermostat appliances by just talking to the device. By using basic utterances we are able to manipulate the devices.

Conversational AI


It is very interesting to have a robot or a chatbot respond to our queries. The chatbot emulates a conversation very similar to a human and it can be deployed across different channels like web chat interfaces, emails, social media platforms or voice-enabled services through mobile. For example, if there is a scenario where one of our appliances has broken down in the middle of the night or at a time wherein we are in need of a fault repair service, it is easier to have a system wherein it takes in the details of the fault and identifies the situation and makes a booking for the service person to make the rectification. The full fault report can be automated and can be made more efficient. The conversation is made through a set of questions and answers and based on these answers, appropriate action is taken by a chatbot. Azure chatbot service is a tool that makes use of a cloud-based platform for developing and managing bots. It can be used to develop bots across different platforms. Integration of the language understanding, the question and answer maker turns the overall bot experience better.

This article is just the tip of the iceberg and just touches an overall view of artificial intelligence. Artificial intelligence and Azure AI has a lot more capabilities and services with which intelligent systems can be built. For a deeper understanding of the artificial intelligence services and azure AI, many certifications and learning tools are available on the Microsoft platform. We can go through the courses prescribed there and take certification exams to improve their understanding of the Azure AI.


Ignite 2020-

Related Articles

Get a free quote for your next project

We can help you build cloud and data products faster. Our Azure certified professionals can help you in driving business value with exceptional quality.

Contact us