Basic Information

  • Project Title / Domain: 2D Unknown View Tomography (Similar to CryoEM)
  • Name: Shreyas Grampurohit
  • Guide: Prof. Satish Mulleti and Prof. Ajit Rajwade

Description of Abstract / Project

This project aimed to develop a new algorithm for reconstructing a detailed 2D image from a dataset where the individual views had unknown orientations and alignments. The proposed method was designed to simultaneously estimate the correct viewing angle and positional shift for each view, with the goal of overcoming the severe image degradation that these uncertainties cause. This 2D investigation was undertaken as a foundational step, as the 3D version of this problem is directly applicable to the cryo-electron microscopy reconstruction challenge.

Whom did you work with?

The prof directly

Relevant Courses

CS754, EE229

How did you connect with the prof?

I emailed Prof. Mulleti, and he set up a time to discuss about a project topic. He talked about the project domain (which was introduced in CS754), gave a very rough idea of what kind of problem in that domain we will be solving and said that we’ll be working with Prof. Rajwade. There was no fixed project problem statement – we finalised a suitable project topic over the course of approximately 3 weeks, after realising that the original problem statement we started out with wasn’t suitable. The final deliverable turned out to be very different from what we had expected in the beginning.

Expectations from Prof

I had discussed with Prof. Mulleti his expectations from me. He said that he expected me to “be an expert in the field” before the SRE ends. Prof. Rajwade was pretty chill regarding the deliverables.

Logistics of the Project

All three of us – Prof. Mulleti, Prof. Rajwade and I – met once a week for approximately 45 minutes. There was no particular milestone framework. The professors expected a significant amount of work to be done before each meeting.

Reading Material

The reading was often papers related to the stage of the project we were at, at that point of time. Prof. Mulleti had suggested a book to read, but didn’t demand on that since it was pretty unrelated to the project.

Tools / Simulations / Softwares / Hardwares Used

No particular specialised software/hardware. Just one programming language.

Outcomes / Deliverables

I hadn’t completed the implementation of the algorithm completely before the SRE presentation deadline. The algorithm had been developed but the specifics of what had been planned was impractical to be directly implemented. The outcome was essentially an algorithm which was not yet entirely implemented and tested. The process of coming up with this algorithm proved the infeasibility of application of various other methodologies.

Extension to DDP

Yes. Not all parts of my SRE were fun and easy. At one point I was stuck at a problem which was very annoying and sounded too “silly” to be a problem. I had spent time trying to understand the root cause of the problem occurring but it looked like the sound of it was too silly to convince anyone that it existed. I had a particular interest in solving Signal Processing problems using modern ML approaches, and that is the core reason that motivated me to continue with the same guides. I had also heard only good reviews from my seniors about the guides, so I decided to continue. Now that I have a lot more experience with them, I can say that I do not regret my decision of extending this SRE to DDP.

Type of Project

nan

Connection to DDP

nan

Advice

Be in constant touch with the guides. Keep a track of your progress and always keep it visible to the guides.