SAMtasia: A Transformer-based Pipeline for Automatic Data Acquisition

Pablo Curiel

University of California, Merced

"SAMtasia: A Transformer-based Pipeline for Automatic Data Acquisition"

As temperatures rise, photosymbiotic marine species are presented with unique challenges. Exaptasia diaphana is a highly adaptable model organism for studying these challenges. Current methods for acquiring experimental data are expensive and require sacrificing the organism. This work focuses on the development of computational tools that will reduce cost and automate the acquisition of data. A pipeline consisting of a convolutional neural network and a transformer-based model is used to accomplish this task. Given input images of aiptasia colonies, this pipeline automatically produces accurate segmentations of aiptasia that can be used for experimental data acquisition (e.g. obtaining counts, measuring oral disk size and color information, etc.).

---