Table 1 Characteristics of publicly- available datasets used for fine-tuning and evaluation of the SAM 2 model, including dataset source, number of unique images, segmentation classes (test split), and annotated masks.

From: A fine-tuned foundational model SurgiSAM2 for surgical video anatomy segmentation and detection

Dataset paper author (citation)

Abbreviated dataset name

Number of unique images

Number of classes (in test split)

Number of masks

W.Y. Hong et al.22

CholecSeg8k

8,080

12 (7)*

45,635

Mathias Carstens et al.23

Dresden

2,431

11 (11)

13,138

Norbert Serban et al.24

UreterUD

586

3 (3)

648

A Murali et al.25

Endoscapes

493

6 (6)

1911

Salman Maqbool et al.26

m2caiSeg

299

17 (17)

2044

  1. *Certain classes were excluded from validation and test subsets due to the inability to split the dataset in a patient-wise manner, which was necessary to maintain generalizability.
Back to article page