- Imageomics, a new field emerging from the convergence of machine learning and computer vision techniques, may be an important tool for addressing questions about the biology of organisms worldwide.
- In this way, research that takes a long time to be done manually can gain significant speed and efficiency.
- Developing the algorithm was inspired by how biologists and ecologists look for features to distinguish various types of biological organisms.
Imageomics, a new field emerging from the convergence of machine learning and computer vision techniques, may be an important tool for addressing questions about the biology of organisms worldwide. Wei-Lun Chao, a researcher at Ohio State University’s Imageomics Institute and a computer scientist at Ohio State, gave a detailed presentation on recent research advances in the field at the annual meeting of the American Association for the Advancement of Science last month. This presentation was part of a session titled “Imageomics: Powering Machine Learning to Understand Biological Features.”
Stating that some research problems can take years or decades to solve manually, imageomics researchers suggest that with the help of machine and computer vision techniques such as pattern recognition and multi-model alignment, the speed and efficiency of next-generation scientific discoveries can increase exponentially.
One way Chao and his colleagues are working to achieve this goal is by creating basic models of imageomics that use data from all types to enable a variety of tasks. Another way is to develop machine learning models that can identify or even discover features to make it easier for computers to identify and classify objects in images. Traditional methods for image classification with feature detection require a large amount of human annotation, but the new method does not. Chao said that in developing the algorithm, they were inspired by how biologists and ecologists look for features to distinguish various types of biological organisms.
Traditional machine learning-based image classifiers achieve a great level of accuracy by analyzing an image as a whole and then labeling a specific category of objects. But Chao’s team is taking a more proactive approach. Their method teaches the algorithm to actively look for features such as colors and patterns in any image that are specific to the class of an object (such as a species of animal) as it is analyzed. The algorithm’s ease of use could potentially allow imageomics to be integrated for a variety of purposes, from climate to materials science. Chao noted that one of the challenging aspects of imageomics research is integrating different parts of different scientific cultures.
Compiled by: Esin Özcan
