Blog

I’ll be introducing the tools and principles we’ll be using for this tutorial, outlining the process at a high level, and then doing a more in-depth walkthrough. Lastly, I’ll go through the processing code in a bit more detail to explain how everything works. And if you’re up for it, I’ll have a little quiz / FAQ at the end. But if you just want to get things up and running quickly, just follow the steps in the Process section. You can get the whole thing running in just a few minutes!

Continue reading Animated StyleGAN image transitions with RunwayML

Human thoughts are persistent, and this enables us to understand patterns, which in turn gives us the ability to predict the next sequence of actions. Your understanding of this article will be based on the previous words that you’ve read. Recurrent Neural Networks replicate this concept.

RNNs are a type of artificial neural network that are able to recognize and predict sequences of data such as text, genomes, handwriting, spoken word, or numerical time series data. They have loops that allow a consistent flow of information and can work on sequences of arbitrary lengths.

Continue reading An Intro Tutorial for Implementing Long Short-Term Memory Networks (LSTM)

Computer-based processing and identification of human voices is known as speech recognition. It can be used to authenticate users in certain systems, as well as provide instructions to smart devices like the Google Assistant, Siri or Cortana.

Essentially, it works by storing a human voice and training an automatic speech recognition system to recognize vocabulary and speech patterns in that voice. In this article, we’ll look at a couple of papers aimed at solving this problem with machine and deep learning.

Continue reading A 2019 Guide for Automatic Speech Recognition

At every point in time during training, a neural network has a certain loss, or error, calculated using a cost function (also referred to as a loss function). This function indicates how ‘wrong’ the network (parameters) is based on the training or validation data. Optimally, the loss would be as low as possible. Unfortunately, cost functions are nonconvex — they don’t just have one minimum, but many, many local minima.

Continue reading An Empirical Comparison of Optimizers for Machine Learning Models

Humans have been trying to augment their reality since the 1990s, when Boeing developed their own AR for assembling their airplanes. The technology gained popularity with the launch of the first version of Google Glass. However, AR has been widely adopted in the consumer market only in the past 5 years, as handheld devices have become powerful enough to host such experiences.

Continue reading Augmented Reality (AR) Development: Tools and Platforms

This is an implementation using the concepts of Q-Learning, which I covered in a previous blog post providing a high-level overview of reinforcement learning (RL).

To help demonstrate these concepts, I’ll be covering how to automate an agent to find the shortest route from its source to a particular destination, recognizing the environment and obstacles, thus learning from its experiences. A typical example would be a robot exploring the environment and finding the optimal way to the goal.

Continue reading Automating an AI to find shortest route using reinforcement learning

Fritz Vision includes a comprehensive API to process pre-recorded video. Some of the things we can do with it include running predictions on every frame, exporting the video as well as extracting individual frames. Fritz Vision also has the power to change old experiences into new ones using machine learning augmentations. In this article, we will see how we can use Fritz to apply filters and perform segmentation of people in a video. This is the result of such an operation.

Continue reading Video Processing in Android with Fritz AI

In a naive encoder-decoder model, one RNN unit reads a sentence, and the other one outputs a sentence, as in machine translation.

But what can be done to improve this model’s performance? Here, we’ll explore a modification to this encoder-decoder mechanism, commonly known as an attention model.

In machine translation, we’re feeding our input into the encoder (green part) of the network, with the output coming from the decoder (purple part) of the network, as depicted above.

Continue reading Attention Model in an Encoder-Decoder

🔋 Low-power consumption is indispensable for autonomous/unmanned vehicles and IoT (Internet of Things) devices and appliances. In order to develop deep learning inference applications at the edge, we can use Intel’s energy-efficient and low-cost Movidius USB stick!

💎 The Movidius Neural Compute Stick (NCS) is produced by Intel and can be run without an Internet connection. The Movidius NCS’ compute capability comes from its Myriad 2 VPU (Vision Processing Unit).

Continue reading A Brief Guide to the Intel Movidius Neural Compute Stick with Raspberry Pi 3

Support vector regression (SVR) is a statistical method that examines the linear relationship between two continuous variables.

In regression problems, we generally try to find a line that best fits the data provided. The equation of the line in its simplest form is described as below y=mx +c

In the case of regression using a support vector machine, we do something similar but with a slight change. Here we define a small error value e (error = prediction – actual).

Continue reading Support Vector Regression in Python Using Scikit-Learn