Table 2 LodeSTAR outperforms other selft-supervised and low-shot methods

From: Single-shot self-supervised object detection in microscopy

 

Dataset 1

Dataset 2

Dataset 3

SoCo

0.85

0.84

0.48

FSDet

0.73

0.75

0.59

InstanceLoc

0.82

0.67

0.43

DETReg

0.71

0.61

0.44

LodeSTAR

0.98

0.97

0.95

  1. F1-score of self-supervised and low-shot object detection methods evaluated on the datasets in the cell-tracking challenge BF-C2DL-HSC (dataset 1, Fig. Fig. 3f), Fluo-C2DL-Huh7 (dataset 2, Fig. 3g), and PhC-C2DL-PSC (dataset 3, Fig. 3h)12. The best scores are highlight in bold. LodeSTAR significantly outperforms the other methods in each experiment.
Back to article page