Important Applications of deep learning.

                                      In this blog, we will study some important applications of deep learning. Deep learning has a wide variety of applications and it can be used in a lot of fields. 

Deep learning algorithms like conventional neural networks (CNN), recurrent neural networks (RNN) and many more help us in a lot of fields.

If you don't know what a conventional neural network and recurrent neural network mean, then click here to know the basics of deep learning and its types.

Table of contents

Applications of deep learning.

1. earthquake prediction using deep learning.

       a. most affected country by                 earthquake       

2.  automatic music generation using    deep learning.                                       

 a. Wavenet.

 b. LSTM (long short term memory)

3. colourization of black and white images.

1. Earthquake.

                        

                    A lot of countries are affected by earthquakes every year. Many people lost their lives due to the earthquake.

How earthquakes are measured.

                                                   Earthquakes are measured in magnitude. A Richter scale is used to check the magnitude of earthquakes. This scale was developed by an American seismologist Charles Francis Richter in 1935.

                             In 1556, Shaanxi, China was struck by the deadliest earthquake in history. It is estimated that around 830,000 people died due to this  earthquake. The magnitude of this earthquake is estimated to be 8.

                                                        Countries like Japan are most affected by earthquakes due to highly active seismic areas. Every year around 1500 earthquakes strikes Japan.

 The worst or deadliest earthquake in Japan was the great kanto which killed around 100,000 people in 1923.

                                                                 Did you know the largest earthquake ever recorded in human history? The largest earthquake of magnitude ever recorded was 9.5 in 1960 in Valdivia, Chile.

Can earthquakes be predicted?

                                                              What if we can predict the earthquakes? Can we predict the exact date, time and magnitude of an earthquake? If we can predict the earthquake it will save a lot of lives.

                                     Yes, it is possible to predict the earthquake using deep learning methods, scientists require huge amounts of data for predicting the earthquake using deep learning.

          A lot of factors are considered while predicting the earthquake like the changes in ground water level, changes in cloud, changes in electromagnetic emissions and many more.

          Although, deep learning is not fully developed to predict the exact date, time and magnitude of an earthquake.

2. Automatic music generation using deep learning.

                           I guess a lot of people love listening to music everyday. Music improves focus while doing work and it improves mood, Music is related to emotions.

 Maybe a lot of people dreamt of becoming a musician or a composer but due to some issues or lack of skills they have given up on their dreams.

 But, in the future it is possible to be a musician by using deep learning techniques.

                        Automatic music or audio can be generated by using two methods. First one is wave net and the other one is LSTM ( long short term memory).

Wavenet.

                        Automatic raw audio can be generated by wavenet. It is a type of deep neural network which is used for generating music.

 Wavenet is able to generate realistic human voice, for doing this it needs to be trained with the recordings of real human speech.

Example : if Wavenet is trained with a musician's voice, it is able to generate similar types of music or audio

LSTM 

                    LSTM full form is long short term memory. It is one type of recurrent neural network (RNN). LSTM is used for a wide variety of applications such as such recognition, handwriting or text recognition and many more.

3. colourization of black and white images.

                                 Three colourization of image was first started by lumiere brothers named as auguste Marie Louis Nicolas and Louis Jean, lumiere brothers invented lumiere autochrome in 1907, the lumiere autoChrome is used for photographic color process.

                                                             More, what about adding colours to old black and white images, normally in the past it was done by human effort. But, now adding colors to image or colourization of image is done by deep learning, which colors images similar to humans.

 Conventional neural network (CNN) is trained for colourization of images. CNN is also used for colourization of black and white images.

Summary

                                                                      In this blog, we have studied some important applications of deep learning and how deep learning algorithms are used in all of the applications.

We obtained some information about the earthquake and use of the Richter scale. 

                       Next, we have studied how Wavenet and LSTM is used for automatic music generation and the last one is colorization of black and white images using deep learning.