[JdeRobot] Fwd: Robotics Academy Project for GSoc 2020

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

[JdeRobot] Fwd: Robotics Academy Project for GSoc 2020

Eduardo Perdices-2

---------- Forwarded message ---------
From: Naman Jain <[hidden email]>
Date: Tue, Mar 10, 2020, 08:57
Subject: Re: Robotics Academy Project for GSoc 2020
To: <[hidden email]>, <[hidden email]>


Hi,
I contacted you two weeks ago for queries about the computer vision GSoC project. I understand you might have missed the mails due to your busy schedule, I thought it best to remind you.
Specifically, my queries were:

Queries: I had some queries about the end goals of the project. Particularly, I am interested in how are the exercises to be integrated with ROS and cameras. While computer vision exercise can be performed solely on offline videos, I understand that using webcam input would be exciting and add more interest to them. So I wanted to ask, what kind of integration do you plan to use between camera input and exercises. For some exercises I was thinking about, it might not be possible to have real-time algorithms too. 

Secondly, in one of the computer vision exercise on FollowFace, you have an explicit requirement of Sony Evi dp100 camera. Would there be other hardware requirements for working on the project? 


I also had some mentioned project ideas, which I would be glad if you could comment on. 
Apart from some simple and fundamental exercises on image gradients and using different filters, corner detection we could add interesting exercises such as panorama stitching, image alignment, noise removal, image inpainting, cartoonifying images (using edge detection and noise addition), style transfer (based on image processing techniques). We could also operate on video tasks using KLT Tracking, video stabilization, optical flow etc.

Thanking you in advance,
Naman
 

On Mon, Feb 24, 2020 at 2:57 PM Naman Jain <[hidden email]> wrote:

Hi,

I am a final year undergraduate studying computer science at IIT Bombay. I am enthusiastic about research primarily in areas of machine learning applied to computer vision. I will be continuing to take graduate studies in computer vision starting Fall'20. 


Motivation: I was looking at GSoC 2020 projects and came across the Robotics Academy project for adding new computer vision exercises and found it matching my interests. I believe with my strong background in computer vision, I would be a good fit for this position and it would also be a nourishing opportunity for me. 


Previous Open Source Work: My work on human pose estimation has been used in several people around the globe. I have also contributed to Pytorch, a deep learning library and committed the `randint` function along with relevant documentation and tests. You can find my other work on my github page.


Queries: I had some queries about the end goals of the project. Particularly, I am interested in how are the exercises to be integrated with ROS and cameras. While computer vision exercise can be performed solely on offline videos, I understand that using webcam input would be exciting and add more interest to them. So I wanted to ask, what kind of integration do you plan to use between camera input and exercises. For some exercises I was thinking about, it might not be possible to have real-time algorithms too. 

Secondly, in one of the computer vision exercise on FollowFace, you have an explicit requirement of Sony Evi dp100 camera. Would there be other hardware requirements for working on the project? 


Ideas: I was thinking of possible exercises to add and I was able to come up with many interesting ideas for the exercises based on projects I have completed or seen. We could add exercises on panorama stitching, image alignment, noise removal, image inpainting, cartoonifying images (using edge detection and noise addition), style transfer (based on image processing techniques). We can also come up with exercises on video (KLT tracking, video stabilization), etc. 

I would think of more cases that can be used to teach computer vision. Please let me know what do you think about these ideas and if I should explore some particular direction you might have in mind. 


Looking towards a positive response.


Thanking You,

Naman Jain

Final Year UG

CSE, IIT Bombay


Web: http://naman-ntc.github.io/

Mail: [hidden email]


_______________________________________________
Jde-developers mailing list
[hidden email]
https://gsyc.urjc.es/cgi-bin/mailman/listinfo/jde-developers