.png){kind=icon}
.png){kind=icon}
Return
FKDNuSeg: Flawless knowledge distillation for lightweight and fast nuclei instance segmentation and classification
.png){kind=icon}
PRE
AI
B
J
J
T
Z
G
Y
C
DOI：10.1016/j.media.2026.104089.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• We develop FKDNuSeg, a lightweight nuclei segmentation and classification model. Only with 4.11M parameters. • FKDNuSeg maintain comparable performance to larger nuclei segmentation and classification models. • FKDNuSeg requires only 9 min and 15 s to complete model inference on a 70,984×80,104 WSI and enabling completion of the entire nuclei segmentation workflow within 30 min (including the post-processing step). • We propose Flawless Knowledge Distillation (FKD), which enhances the classification capabilities of model.
Keywords:
FKDNuSeg
lightweight model
nuclei segmentation
knowledge distillation
fast inference
Journal
IF:
11.8
Papers:
3.7K
Citations:
2.4W
Organization
Cited Papers