GSoC Project Idea 16.2: Semantic Image Processing for Developmental Data Science

This project is a follow-up to an OpenWorm ( Foundation-sponsored GSoC project from 2017 (link: In the 2017 project, former GSoC student Siddharth Yadav ( developed a plugin for the ImageJ ( package that provides our community with a quick method for segmenting raw image data. In this project, we hope to expand upon the communities’ capacity to analyze images acquired via microscopy. In the process, you will also help to establish a training dataset composed specifically of developmental phenotypes such as embryos and emerging neural systems.

Aims: The goal of this project will be to create a semantic segmentation library that can be used to either augment directly or complement the ImageJ plugin. Rather than a easy-to-use tool that is geared towards the analysis of developmental data, the next step will involve refining the feature identification capacity of our segmented images. The applicant will work with the DevoWorm ( group to determine a feature space relevant to a specific set of research questions and potential applications. The analytical approach can take one of two directions: Semantic and Instance Segmentation or Region Proposal Networks/Fast Convolutional Neural Networks. Both of these approaches are state-of-the-art methods in this area, and other approaches can be proposed in consultation with the mentor. This experience would provide the applicant with a solid research experience in applied algorithmics and data science.

Skills: open source development experience with Python. Some prior experience in data science and/or machine learning would be helpful. An interest in the underlying biological processes is essential. Applicants with good communication skills and familiarity with open science practices are expected.

Mentor: Bradly Alicea (, Stephen Larson (


I’m Vinay Varma, Junior CS student at Amrita Vishwa Vidyapeetham, Amritapuri, India. My career interest lies in the field of Computer Vision and Data Science. I have some academic and professional internship experiences involving the same. I’m exactly looking for something which would give me solid research experience involving Image Processing and also touches the medical field up to a great extent.

The existing work which was done in GSoC’17 is good and I’m currently diving into the codebase. Apart from the standard Image Processing techniques (boundary detection), there has been a lot of advancement in the semantic segmentation and feature extraction (images) fields ( Deep Learning approaches) since the past two years, which means there is a lot to explore. I’m also on a lookout on how this can be extended to extract useful features from segmented images. I think that figuring out how we can derive some sense and make use of these segmented images will be a good starting point. I’m looking forward to discussing the pros and cons of various approaches with the mentor.