Introduction to AR Core

What is AR (Augmented Reality)

• Blend real world in to the digital world

• Integrate virtual content with the real world as seen through phone's camera

ARCore is Google’s platform for building augmented reality experiences
 

What is Google’s Tango Project (Prior to AR Core)

        ARCore has its origins in Tango, which is/was a more advanced AR toolkit that used special sensors built into consumer mobile device.
        Google’s (Peanut phone , Yellow Stone tablet) , Intel’s Real sense Smart phone , Lenovo Phab -2 , Asus Zen phone are equipped  with special sensors
        Later Google stopped Tango project as ARCore is evolved with less dependent on sensors as compared to Tango

Hardware of Phone used by ARCore

• Fundamental concepts behind ARCore

• Motion tracking

• Feature points (through camera) + Motion sensors (through IMU: Accelero , Gyro)

• Point , PointCloud , Pose are some of the classes exposed by ARCore SDK we can see how  user's position is tracked in relation to the feature points identified on the real couch.

• Environmental understanding 

• Uses technique called meshing 

• Cluster of feature points used and returned as planes to applications along with each plane’s boundary

How does AR Core works

AR Core fundamentally does two things

    1. Motion Tracking
        Building it’s own understanding of real world (Environmental understanding , Light Estimation , Orientation points , Anchors , Trackables

    2. Environmental limitations
        For now, limitations that may hinder accurate understanding of surfaces include:

• Flat surfaces without texture, such as a white desk

• Environments with dim lighting

• Extremely bright environments

• Transparent or reflective surfaces like glass

• Dynamic or moving surfaces, such as blades of grass or ripples in water

        When users encounter environmental limitations, indicate what went wrong and point them in the right direction.   

            AR application that has identified a real-world surface through meshing. The plane is identified by the white dots. In the background, we can see how the user has already placed various virtual objects on the surface.

Light estimation

        Average intensity and color correction of a given camera image
Scene of realism is increased by applying the same light condition to virtual objects LightEstimate , LightEstimate.State are some of the classes used to get lightning condition User Interaction , Orientation points , Anchors and Trackables.
        Takes an (x,y) coordinate corresponding to the phone's screen
Projects ray into camera’s view of the real world
        Returns Plane/feature points that the ray intersects and Pose (kind of OpenGL model matrices) of that intersection in world space
HitResult , Pose , Anchor (fixed location and orientation of real world) are some of the classes provided by ARCore in this context
Trackable (interface) is something that ARCore can track and that Anchors can be attached to 

Note: 
        Because ARCore uses clusters of feature points to detect the surface's angle, surfaces without texture, such as a white wall, may not be detected properly.
        To reduce CPU costs, reuse anchors when possible and detach anchors that you no longer need.

Link for reference: https://developers.google.com/ar/develop/java/cloud-anchors/overview-android

References:
Samples: https://github.com/google-ar/sceneform-android-sdk.git 
https://developers.google.com/ar/develop/java/quickstart

Books on ARCore

Did you find our blog helpful ? Share in comment section. Feel free to share on Twitter or Facebook by using the share buttons

Read More

Android A11 features

                       

             Android 11 introduces great new features and APIs for developers. Android 11 focus more on People-centric and expressive, with a new controls space and more privacy features.
            Extend your apps with conversation notifications and bubbles, try one-time permissions, surface devices and media in the controls. Work faster with tools like compatibility toggles, ADB incremental installs, and more!

Lets look into Android 11 features

Conversation notifications

                There are now three notification categories: Conversations, Alerting, and Silent. The Conversations section, quite obviously, houses all your conversations. This would mean any app where you are directly communicating with someone else, including text messages and chat apps.     

            Meanwhile, the Alerting and Silent sections act as they have before in Android 10. You can also easily silence notifications from certain apps, which would push all future notifications to the Silent section. With Android 11, you now have more control over notifications than you ever had previously.

Notification history

            In Android 11! A new feature gives you the option of saving every single notification that landed on your phone over the past 24 hours. You can check the running list, find the notification you accidentally swiped, and see what you missed.

Related: Here’s a list of Android 11 phones

Chat bubbles in Android 11

             For all chat applications Android, you already know how chat bubbles work. With Messenger, a “chat head” appears on your phone that overlays on top of pretty much every other app. A quick tap of the icon launches the chat and then you can minimize the chat back to an icon. Done with the conversation? You can remove the chat head until the next conversation starts

Android 11 screen recorder

            The screen recorder function lives in the Quick Settings tiles. You tap the Screen Record feature which gives you a few options before you start recording. For example, you can choose whether or not your screen-taps should also be recorded and whether the phone should capture audio, too.

Media controls

                If you are playing music on your Android 10 phone, a music player appears at the top of your notifications drawer. Of course, with Android 11, that section of the drawer is now reserved for conversations, so the media player needed to move. Google decided to move it one rung up to the Quick Settings section.    
                When you swipe down your notification drawer, the media controller will be pretty small. It will show you the app it’s related to, cover art, basic controls, and on which system the media is playing. If you pull down again on the drawer, the alert expands and shows the information

Smart device controls

                Google added a new section in Android 11 that allows you to easily control your various devices without needing to open an app.

                You can hold down the power button to launch the new tool. At the top, you’ll find the usual power features, but underneath, you’ll see a lot more options. There’s a Google Pay shortcut that allows you to quickly choose which payment method you want your next contactless transaction to use. Under that, you’ll see a bunch of buttons connected to your various smart home products.
                
        The Device Controls feature, available starting in Android R, allows the user to quickly view and control external devices such as lights, thermostats, and cameras from the Android power menu. Device aggregators (for example, Google Home) and third-party vendor apps can provide devices for display in this space.​

One-time permissions and auto-reset

            Android 11 gives the user even more control by allowing them to give permissions only for that specific session.
            Similarly, Android 11 will now “auto-reset” apps you haven’t used in a while. If you granted location data permissions to an app that you haven’t opened up in a long time, Android will now revoke all permissions. Next time you open the app, you’ll need to approve those permissions again. 

Dark Theme 

            With Android 11, users can now schedule the dark theme using one of two different metrics. You can schedule a dark theme to turn on or off when the sun sets or rises. You can also set up a custom schedule for dark mode activation if you wish.

 

Some of the other features are 

Concurrent Camera Streaming : Many devices have the hardware capability to stream multiple camera sensors concurrently.

Camera Bokeh : to make the bokeh feature available to third-party apps​

Generic Kernel Image​ : Android R introduces the Generic Kernel Image (GKI), which addresses kernel fragmentation by unifying the core kernel and moving SoC and board support out of the core kernel into loadable modules.​

Device Type Limit​ : In Android R, no limit on the number of audio device types to allow new audio device types to be added​

Soft restarts​ : Android R supports soft restarts, which are runtime restarts of processes in the user space used to apply updates that require a reboot (for example, updates to APEX packages).​

Zombie permissions : Android will automatically revoke an app’s permissions after an undetermined time of inactivity—somewhere between 60 and 90 days. Launching an app within that time reinstates any permissions you’ve granted​

Notification/Sound Muting​ : Device vibration and sounds can hurt image and video quality, especially for cameras with optical image stabilization module​. camera app could use the (DnD) APIs to mute the vibration and sounds
Offline Processing​: Today camera apps must wait for all requests to be fully processed before it can disconnect from the camera or switch to a different configuration, The offline processing API allows the camera HAL to perform post-processing in the background to improve the latency of mode switching or camera closing​

Android 11 updates via Play Store

            Each year, Google releases the latest version of Android. Each month, it pushes out the latest Android security patch. Both of these updates get funneled to your phone either by your carrier or equipment manufacturer. Because of this, some phones get many updates very quickly, while others either get them much slower or not at all.o counteract this, Android 11 gives more power related to updates over to the Google Play Store. This allows Google to bypass carriers and OEMs entirely and push out updates to everyone.

App-pinning to the share sheet

            You can now pin apps to your share sheet in order to easily access them whenever you want to share something.This feature allows you, the user, to control which apps appear at the top of the list when you want to share something.

Improved prediction tools

            Android 11 will seemingly reduce the work you need to do on your phone, by predicting your habits and patterns.
            One such example of this is smart folders, so you can let Android 11 automatically sort your apps into folders of similar apps, like games or productivity tools.
            App suggestions is also tweaked to suggest apps based on your routine - for example, if you always log onto your Fitbit app first thing in the morning to examine your sleep habits, the phone will now automatically pop that app into the Home screen in the morning so it's easily accessible.
            Finally, apparently the Smart Reply feature already usable in Android phones has received some tweaks. This mode suggests some automatic responses when you receive a message, letting you reply with one tap (if any of the responses are appropriate) but it's not clear what's new here.

New phones coming with Android 11

• OnePlus 8

• Oppo Find X2 and Find X2 Pro

• Realme X50 Pro

• Xiaomi Mi 10 and Mi 10 Pro

Did you find our blog helpful ? Share in comment section. Feel free to share on Twitter or Facebook by using the share buttons

Read More

Best Machine learning books

        

                  Machine learning and artificial intelligence are growing fields and growing topics of study.  In this article, we’ll review some of the most popular resources for machine learning beginners. Some of these books will require familiarity with some coding languages and math.

1. Machine Learning For Dummies” by John Paul Mueller and Luca Massaron

Topics Covered :

• Introducing How Machines Learn : the Real Story about AI, Learning in the Age of Big Data, Glance at the Future

• Preparing Your Learning Tools : Installing an R, Python , Coding in R Using RStudio, Python Using Anaconda and exploring other machine learning tools

• Getting Started with the Math Basics : Math Behind Machine Learning, Probabilities, Statistics,Cost Functions, Error Curve , Greedy Classification Trees, Incredible Perceptron, Greedy Classification Trees. Validating Machine Learning : Training, Validating, and Testing

• Learning from Smart and Big Data : Preprocessing Data, Leveraging Similarity , Working with Linear Models, Neural Networks, Support Vector Machines

• Applying Learning to Real Problems : Classifying Images, Scoring Opinions and Sentiments, Recommending Products and Movies

Where to buy: Amazon

2. Programming Collective Intelligence” by Toby Segaran

Topics Covered :

• Introduction to Collective Intelligence, Making Recommendations

• Discovering Groups : Supervised versus Unsupervised Learning, Word Vectors, Hierarchical Clustering, K-Means Clustering, Viewing Data in Two Dimensions

• Searching and Ranking, Optimization

• Document Filtering : Filtering Spam, A Naïve Classifier, The Fisher Method, Modeling with Decision Trees

• Building Price Models :k-Nearest Neighbors ,Weighted Neighbors

• Advanced Classification: Understanding Kernel Methods and SVMs , LIBSVM , Finding Independent Features

• Algorithm Summary : Bayesian Classifier, Decision Tree Classifier, Neural Networks, Support-Vector Machines, k-Nearest Neighbors, Clustering, Multidimensional Scaling, Non-Negative Matrix Factorization, Optimization

• Different third party libraries and Mathematical Formulas

Where to buy: Amazon

3. Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow

Topics Covered :

• What Machine Learning is, what problems it tries to solve, and the main categories and fundamental concepts of its systems, Optimizing a cost function, Handling, cleaning, and preparing data, Selecting a model, The challenges of Machine Learning

• The most common learning algorithms: Linear and Polynomial Regression, Logistic Regression, k-Nearest Neighbors, Support Vector Machines, Decision Trees, Random Forests, and Ensemble methods

• unsupervised learning techniques, including clustering, density estimation, and anomaly detection

• What neural nets are and what they’re good for, Building and training neural nets using TensorFlow and Keras

• Neural net architectures: feedforward neural nets for tabular data, convolutional nets for computer vision, recurrent nets and long short-term memory (LSTM) nets for sequence processing, encoder/decoders and Transformers for natural language processing, autoencoders and generative adversarial networks (GANs) for generative learning

• Techniques for training deep neural nets, Training and deploying TensorFlow models at scale 

Where to buy: Amazon

4. Natural Language Processing with Python

Topics Covered :

• Natural Language Processing with Python and spaCy will show you how to create NLP applications like chatbots, text-condensing scripts, and order-processing tools quickly and easily

• Using Python for Natural Language Processing

• The spaCy Library, How Can Computers Understand Language?, Using Machine Learning for Natural Language Processing

• Neural Network Models, Convolutional Neural Networks for NLP,

• Extracting and Generating Text with Part-of-Speech Tags,, Finding Patterns Based on Linguistic Features

• Implementing and Deploying a Chatbot Works

Where to buy: Amazon

5. Machine Learning in Action

Topics Covered :

• Machine learning basics, How to choose the right algorithm, python, NumPy library, Classifying with k-Nearest Neighbors

• Decision trees, probability theory: naïve Bayes

• Logistic regression, Support vector machines, AdaBoost meta-algorithm, linear regression

• Tree-based regression, Unsupervised learning

• Big data and MapReduce Hadoop Streaming, Machine learning in MapReduce, The Pegasos algorithm

Where to buy: Amazon

6. Machine Learning with TensorFlow

Topics Covered :

• Machine-learning fundamentals, Types of learning : Supervised learning, Unsupervised learning, Reinforcement learning, TensorFlow essentials

• Core learning algorithms Linear regression,Polynomial model, Regularization, Using logistic regression, Multiclass classifier

• Automatically clustering data : K-means clustering, Clustering using a self-organizing map

• Hidden Markov models, The neural network paradigm : Reinforcement learning, Convolutional neural networks, Recurrent neural networks

• Sequence-to-sequence models for chatbots

Where to buy: Amazon

7. Introduction to Machine Learning with Python: A Guide for Data Scientists

Topics Covered :

• Scikit-learn, Jupyter Notebook, NumPy, SciPy, matplotlib, pandas, mglearn

• Supervised Machine Learning Algorithms : k-Nearest Neighbors, Linear Models, Naive Bayes Classifiers, Decision Trees, Kernelized Support Vector Machines, Neural Networks (Deep Learning)

• Unsupervised Learning and Preprocessing :  Dimensionality Reduction, Feature Extraction, and Manifold Learning, Clustering

• Representing Data and Engineering Features, Model Evaluation and Improvement, Algorithm Chains and Pipelines, Working with Text Data

Where to buy: Amazon

Did you find our blog helpful ? Share in comment section. Feel free to share on Twitter or Facebook by using the share buttons

Read More