Self-sufficient prototype of a self-driving car.

---

I have been working on a project to build a prototype of a self-driving car which is self-sufficient and self-capable. The aim is to make a car that can steer itself, detect obstacles and read road signs so as to take corresponding actions.
This is basically a fast, completely pythonic way of doing this using simple Feed Forward Neural Networks.
I built a track on which I train my car.
The car has a Raspberry Pi 2 mounted on top of it. A PiCamera and USB WIFI module are connected to it. The Pi creates a ad-hoc network with which one can access it.
While training the car to take decisions, I collect a sample to 10,000 images of the track of dimensions 150pi*75pi each, with their corresponding movement direction. Subsequently, I train a neural network(on my laptop) with a single hidden layer with 100 neurons using these input images and decisions.
The trained network is passed on as an XML file to the Raspberry Pi 2. During runtime, a new image is captures, passed to the neural network which gives a movement direction.
At every instance, the car checks for obstacles using an HC-SR04 ultrasonic sensor. With the distance formula, the distance of obstacle is calculated and car stops if obstacle is closer.
For each image frame captured, the car checks for road signs. A haar cascade classifier can be trained for each road sign. If found, its distance is calculated. If the sign is within range, action corresponding to the sign is taken.
The car has onboard LEDs for user viewing. Also the Pi hosts a web page that shows the image stream and decision taken.
The car is properly powered which gives it a high runtime.
While running, the car doesn't need any external computational power. Hence it is self-sufficient.
For details, please have a look at the video below. I hope you like the project.
I have uploaded its video on YouTube.



The source code is available on GitHub.
AutonomousCar

• Date: 1st January 2016 - 29th February 2016

Close

Hand gesture based Car Movement

---

I have been working on a model of a car that could run based on the driver's hand gestures.
Check the video below.





The source code is available on GitHub.
Hand Gesture based Car Movement

• Date: March 2016

Close

Inline Arduino Skating

---

Featured on 'HACKADAY' : KNOW YOUR SPEED ON ROLLERBLADES

Recently I bought my new pair of inline skates. And I have modified it a bit using Arduino, DHT11, JHD 162A, HC SR04 etc.
I am trying to measure the rpm, in turn, the linear velocity of my wheels while I am skating. I have used an ultrasonic sensor, with the help of which I am measuring the speed of my skates only when they are on the ground(about 15 cm). I am also measuring the temperature and humidity of the surrounding. All the values are displayed on the 16*2 LCD display.
I have uploaded its video on YouTube.See the videos below





The source code is available on GitHub.
Inline Arduino Skates GitHub Code

I would like to thank my senior Chirag Nagpal for giving me guidance.

• Date: 8th February 2015

Close

SoundCloud User Classification using ambient metadata.

---

Will be updated soon!

• Date: --/--/----

Close

ThankYou Facebook

---

Many people wish you 'Happy birthday', on your birthday, on Facebook. Are you manually going to like and comment? Here's a cool shortcut. This is basically a script to thank all the people who post 'Happy Birthday' on your timeline.
The code can be found at GitHub.
ThankYou Facebook.

• Date: 20th November 2014

Close

Internet Speed Plotter

---

One day, the internet speed of my college was fluctuating. I wasn't able to do anything. So, I thought of making a script to observe the internet speed of my college over a period of a one day(24 hrs) and plotted a graph.
The result was pretty interesting. I observed that the sweet spot to use internet is the night-time i.e. between 2am - 4am.

The source code can be found on GitHub.
Internet Speed Plotter

I would like to thank my senior Chirag nagpal for the guidance.

• Date: 21st November 2014

Close