Eric and Wendy Schmidt Center Funded Research and Innovation Scholar
Evaluation of Neural Networks for Object Segmentation in Biological Images
- Computer Graphics and Vision
Many researchers can gain valuable insights into their research using AI models but only few are familiar with their inner workings to use them efficiently. This presents a usability problem that could perhaps be solved with a decent easy-to-understand interface, but training and inference, for large models, tend to be compute-heavy.
This makes it difficult for the biologist, our target user, to use these progressive tools. We propose Piximi, an object segmentation tool for biological images that would bring an intuitive interface coupled with models running on the browser without a server this would mean efficient transfer learning and machine learning workflows that allow for varied model adaptability, without leaving the browser.
I am interested in the dual areas of computer systems & optimization and machine learning. This project exists in this beautiful intersection and I believe that by bringing my joint experience in both fields, I might have interesting ideas that will take us steps closer to our goal. I am also motivated to make this work since its very important in democratizing artificial intelligence.