Adapting Video Diffusion Models for Time-Lapse Microscopy

Alexander Holmberg, Nils Mechtel, Wei Ouyang · 2025

Read Adapting Video Diffusion Models for Time-Lapse Microscopy by Alexander Holmberg, Nils Mechtel, Wei Ouyang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-18583.txt.

Description

Alexander Holmberg, Nils Mechtel, and Wei Ouyang present a 2025 paper on generating fake time-lapse microscope videos of living cells. Video diffusion models make striking natural footage, but they have barely been tried in microscopy. Real cell videos are costly to record, and synthetic ones could aid training and testing.

The team fine-tuned a pretrained video model on sequences of HeLa cells dividing. They tried three ways to steer the output: text prompts built from measured traits, direct number embeddings of those traits, and extending a real first frame. Scores for shape, growth, and movement showed fine-tuning greatly improved realism. The model even stayed coherent past its training length. Precise control of traits, though, stayed hard.

Catalog record

CREATORAlexander Holmberg, Nils Mechtel, Wei Ouyang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE37 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-18583

Browse the full library · About the Tower of Babel