-
Notifications
You must be signed in to change notification settings - Fork 1
/
midsem.txt
38 lines (30 loc) · 2.1 KB
/
midsem.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
# Introduction
1. Brief About Title
Our project aims to develop a computer program that automatically detects and extracts geometry of craters and sand dunes on
mars and the moon.
2. Application
There are multiple applications and the most important one being to identify safe landing site for rovers or manned missions.
3. Necessity or Why is it required?
Crater and sand dune detection was largely done manually which is a time consuming endeavor, difficult task and subject to
human errors. Automation would greatly speed up the process and help increase the efficiency of the process.
# 2nd Approach
We shifted focus to an increasingly popular methodology called machine learning and more specifically deep learning.
These methodologies show better results and increased efficiency which was never seen before.
Fundamentals of Deep Learning -
It is essentially an algorithm that tries to emulate the way the human brain learns and works.
1. What?
Specifically we will be using convolutional neural network. It is the most popular deep learning architecture.
CNN is now the go to model on every image-related problem. One of the main advantage is that it automatically
detects the important features without any human supervision.
2. Technology Used
We will be using tensorflow and keras ML Frameworks with python and utilise GPU accelerated performace
provided by an online solution called Google Collab
3. About CNN
Convolution is a mathematical operation to merge two sets of information.The architecture consists of mainly
The basic architecture contains
* Convolution layers
* Non-Linear or Activation Function which is typically a ReLU rectified linear unit
* Pooling layers - Mostly utilising Max Function
- Shortens Training time and combats overfitting, they downsample each feature map independently
We have gone through various research papers that attempted to do something similar and found ## the most appropraite
and we have decided to follow it closely and implement according to our needs.