UCTLFANet: a low-rank fine-tuning model for microalgae image segmentation
Ziyue Liu, Yuan Cheng, Dan Liu, Meiyi Jiang, Wenhao Yue, Shuo Yan, Fengyu Tian, Liming Cao, Daoming Jia, Yanbo Jiang, Hai Bi · 2026
Read UCTLFANet: a low-rank fine-tuning model for microalgae image segmentation by Ziyue Liu, Yuan Cheng, Dan Liu, Meiyi Jiang, Wenhao Yue, Shuo Yan, Fengyu Tian, Liming Cao, Daoming Jia, Yanbo Jiang, Hai Bi (2026) 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/pmc-13415794.txt.
Description
Liu and co-authors present, in 2026, UCTLFANet, a neural network for segmenting microalgae in microscope images, segmenting meaning outlining each cell. It supports species analysis, cell counting, and health assessment in marine research. Its novelty is a fine-tuning module from large language model practice: LoRA-FA adjusts the network's weights through low-rank updates instead of retraining everything, slashing compute.
The authors inserted the module into different layer types and found the fully connected layers gave the best segmentation outcomes. Head to head, their model outperformed UNet, UNet++, Attention-UNet, and UCTransNet on their microalgae benchmarks, and bolting the module onto those rival networks improved each of them too, though none caught up. The low-rank approach cuts trainable parameters by 3.6 times in method comparisons, which the authors argue makes the tool practical for marine microbial imaging, where labeled data and hardware are both limited.
Catalog record
| CREATOR | Ziyue Liu, Yuan Cheng, Dan Liu, Meiyi Jiang, Wenhao Yue, Shuo Yan, Fengyu Tian, Liming Cao, Daoming Jia, Yanbo Jiang, Hai Bi |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |