Motion segmentation in biological images
The goal of this research is to extract instances of C. elegans (roundworm) in microscopic video, characterize their motion patterns, and classify them into normal and abnormal classes. The student researcher will develop Python code for video segmentation using current-generation neural network methods (e.g., SAM, DINO). Segments corresponding to isolated instances of C. elegans will be tracked, and their patterns of movement and articulation ("squirming") characterized by features such as Fourier moments. These features will be provided to a classifier that learns to distinguish worms with a normal diet from worms with an adulterated diet (narcotics, etc.). Characterization of segmenter and classifier performance will be performed. An early component of this work will be the assembly of a corpus of video from online sources to use in training.
The Computer Vision Research Laboratory is home to three faculty and about tweny graduate students, undergraduate researchers, and postdoctoral scholars. Research teams meet frequently and conduct research in a variety of topic areas relating to the analysis and interpretation of imagery.