Volcanoes On Venus
Welcome to my Project Volcanoes On Venus.
In this project, I have analyzed the images of Volcanoes on Venus captured by NASA’s Magellan spacecraft in 1990.
Data Preparation page explains the analysis and data cleansing I carried out on the input data.
Git Hub page has the code and other documents used for analysis.
Background
The primary objectives of the Magellan mission were to map the surface of Venus with a synthetic aperture radar (SAR) and to determine the topographic relief of the planet.
At the completion of radar mapping 98% of the surface was imaged at resolutions better than 100 m, and many areas were imaged multiple times.
In the analysis of the data captured by the spacecraft they found volcanoes on the surface on Venus, volcanoes that can be used to make a automatic machine that can detect them.
Report
The goal of this project is to experiment with and study the different types of neural network architechtures, from simple hand coded models to complex CNN’s using Tensorflow/Keras, in the process of identifying Volcanoes from a given surface image of Venus.
Here are some fun facts about Venus:
Venus does not have any Moons and so is single.
Venus orbits the Sun every 224.7 Earth days while the rotation period is 243 Earth days. So, you are year old even before you are one day old on Venus.
Models
I borrowed a lot of ideas from the Deep Learning Specialization I completed recently and I implemented the following architectures:
Logistic Regression
Shallow Neural Network
Deep Neural Network
Convolution Neural Network
In this project, I started with implementing a simple hand crafted logistic regression model using Numpy. Later I implemented more sophisticated models using SKLearn and Keras/Tensorflow.
The Logistic Regression models and their results can be found here: Logistic Regression Model
The Neural Net models along with CNN’s and their results can be found here: Neural Network Model
Dataset
The dataset is obtained at Kaggle