Inspiring Deep Learning Applications: Ready-Set-Coding


Also known as artificial neural networks, Deep Learning is a sub-category under the broad machine learning concept that is concerned with algorithms that replicate human behaviour and brain.

Deep Learning is not a new concept as its existence can be traced back to 1943 when Warren McCulloch and Walter Pitts worked on a combination of mathematics and algorithms called “threshold logic” to create a computer model that mimicked human brain. Today, the logic is used extensively in many industries to generate reasoning solutions and create artificial intelligent systems to resolve complex human tasks. Let’s explore applications where deep learning has become a prerequisite to surpass time taking conventional problem-solving methods.

Creating Realistic Videos: ‘Synthesizing Obama’

A team of computer vision researchers under the leadership of Supasorn Suwajanakorn from the University of Washington created a realistic video of former US President Barack Obama from the audio files mined from 14 footages of an interview and a talk show which was telecasted years ago. Among so many personalities and leaders, they chose Obama because machine learning algorithms required extensive readily available videos in the public domain and the internet has enough presidential videos complementing the prerequisite. Further, they added a recurrent neural network to support the system to create facial expressions from the audio clips. The researchers focused on the lower face (mouth and chin), cheeks, and neck, and corrected a little jaw area. The other face specifications like eyes, head, and torso were taken from the stock footage. However, soon experts witnessed that the video lacks sync with the audio and video clips had irregular face structure, but the experiment wasn’t a total failure as image processing tasks enlightened more neural network algorithms.

Atari Game Breakout

Be it a Space Invaders, World of Warcraft or Doom, deep learning communities are roping in to train computer systems to challenge human intelligence in almost every online game. To complement this fact, one of the biggest examples to see today is the Atari game Breakout where Google’s DeepMind researchers used deep reinforcement learning to teach the computer to play this game. No one trained or programmed the system with the game instructions but allowed the system to use keyboard according to its intelligence. Surprisingly, in just two hours, the system became proficient in playing the game.

Photo Recognition

It’s amazing to see how Facebook suggests and tags the names of people who are present in the photo without typing even its initials. Deep learning systems are capable of automatically recognizing and classifying pictures. Another deep learning wonder can be seen in Google photos when it labels your photos with the names and places to give easy identification and accessibility. In research, Timnit Gebru experimented 50 million Google Street View Images with the deep learning system and what he found was unbelievable. The system recognized and categorised over 22 million cars along with their models, makes, mechanisms, and year of the invention.


The language has not remained a barrier anymore with deep learning algorithms as Google Translate app can automatically translate almost every language and images in real-time. Pick any image that has texts written in any different language and see how Google reads the text and translates it in your preferred language.

Self-Driving Cars: From an idea to the road

Let’s walk down in the lane of history when people would only think of self-driving cars in their imaginations. Today, AI systems like deep learning have turned the idea into reality. Companies like Tesla are successfully inventing cars that don’t need human interference to drive the car while detecting and understanding the objects, road signs, and people.

The artificial neural network has fuelled the invention of smart mechanisms in almost every industry. The technology has come a long way and its current stage of deep learning cannot be considered lesser to those in The Matrix and The Terminator. In 2018, Google Cloud released Neural Architecture Search (NAS) with a motive to reduce the necessity of Machine Learning engineers while operating a system. It is way quicker and accurate than conventional methods. The aftermath of 2018 neural network’s big success made it limpid that in the coming years the businesses will witness a revolution with smarter inventions to stun the world with their products and services.