Abhishek shankar's proposal for Github Externship

[abhishek-a05]

Github Externship Proposal

• Student ID: abhishek_a05

• Application No -- 31-05_abh665_tfa_291

• Mentors: Abhishek_ , zeekhuge , jkridner , pratimu

• Repository: https://github.com/abhishek-a05/Latest-Images

Status

This is Abhishek Shankar's proposal for Github Externship.

About you

• IRC: abhi_shek
• School: Manipal Insitute Of Technology
• Country: Mumbai, Maharashtra,India
• Primary language (We have mentors who speak multiple languages): Hindi, English
• Typical work hours (We have mentors in various time zones): 6-8am and 12-2pm or 6-8pm and 9-11pm IST
• Previous externship and open source experience: I don't have any experience with open source projects

About your project

Project name: Real Time Lane Detection

Description

The purpose of this project is to find a feasible solution
to detect road lanes on a low voltage computer that can be easily powered in a regular auto
vehicle. The board can be attached to the vehicles dashboard , and a camera facing the road can capture real time video. The Technologies and languages used are Linux Kernel, Python and C.

Timeline

Provide a development timeline with a milestone week and any pre-work. (A realistic timeline is critical to our selection process.)

Date	Milestone
week 1	Undertanding the RISC-V architecture & studying the board
week 2 & 3	Undertanding the linux kernel implementation of the project
week 4 to 5	Talking to the mentors
week 6 to 11	Developing the model
week 12 & 13	Finalising

Detailed milestone for week 6 to 11:-

• Gathering , cleaning , processing video data
• Choosing a suitable architecture for neural network
• quantizing the model to fit it in the beagle -V starlight
• testing with live camera feed, comparing accuracy , further improvements

Experience and approach

I'm currently pursuing Electronics & Instrumentation in my Pre-final year & pursuing a minor in Embedded systems.I like to work on embedded systems which can incorporate core machine learning/Deep learning techniques. I've previously worked on developing deep learning models for object tracking & detection . I've also worked on Arduino and have developed a mini alarm system , incorporating bluetooth for alerts.

I have a good grip on python , C/C++ , Data structure & algorithms, Linux and deep learning techniques.
I believe i will fair well and i will develop and deploy my solution well within time.

What will you do if you get stuck on your project and your mentor isn’t around?

I will read relevant research papers and material to figure out a solution, search youtube or google. I will also consult my friends who have worked on similar projects and try to solve the problem .

Benefit

Real-time lane detection has wide-spread use in intelligent ground vehicles/robots , that are used in remote locations or places where humans cant access such as a nuclear reactor , or a live fire in a building. By developing this model using the beagle-V we can use it as an off the shelf model for lane detection without having to train the model from scratch, and that too with high accuracy.

Conclusion

I would really like to make this project, with the help of your organisation , and actually make a contribution to open-source and mankind. This project can become more than what i envision with the right direction and guidance.