Lane Detection with Deep Learning
Earned the Best Project of the Week badge on Career Karma
"It is true that human beings can identify lane line markings while driving with basic training,
but sometimes, based on the number of crashes and accidents,
it is understandable that
they can have a disadvantage of not always being attentive.
Although, it is not that easy for computers to learn identifying lane
line markings
but after learning the task, there would not be any distractions
for them and they have this advantages
that the rate for crashes and accidents caused by distraction would be less and computers can take over
this task from human driver. Using deep learning, I have used transfer learning and re-train
a pre-trained model (PINet ).
The model is based on a Convolutional Neural Network (CNN) which perform well on image dataset. CNNs work well with images by looking at them in pixel level."
To see this project
click here
Indexing Natural Disasters and Outbreaks
"The COVID-19 global pandemic has impacted all of our communities.
In some areas, the impact of the pandemic has been exasperated by other natural disasters.
In the state of California, wildfires have been devasating in recent years.
Being part of the ring of fire and laying on top of large number of faults,
earthquakes are also a constant threat. There is a shortage of tools available
for
looking at the combined effect of the pandemic and other natural disasters.
This makes it challenging for decision makers to assess risk and make appropriate prepardness plans.
Here, we will provide a tool to visualize the concurrence of COVID-19 hot spots,
wildfires, and earthquakes in California."
To see this project
click here
Finding your Favorite Reddit Community
"There are about 138,000 active subreddits among a total of 1.2 million subreddits.
But, how is it possible to find your own community with the topics you have interest in?!
Well, that is why I am here! I try to help you out but how?
I collected more than 100,000 comments using webscrapping and APIs from different subreddit categories. By using CountVectorizer and TFIDF and Naive Bayes my Natural Language Processing (NLP) classification model here goes over all the
subreddits that you have some interest in and BAM! It gives you the most key words of each
subreddits and you can easily decide which community is for you! Why am I think my model is good?
Since my model has an accuracy of %99 to identify the certain posts and vocabularies from any subreddits."
To see this project
click here
Predicting Housing Market Prices
"The Ames Housing Data set contains information regarding the houses sold in the years 2006 to 2010 in Ames, Iowa. The main purpose of this project is to model the home prices and I used different techniques using regression
(of course!) and some feature engineering. The final accuracy
of this model is 94%."
To see this project
click here