Table 1 Mean Dice Similarity Score obtained from the different prompt generation techniques on all the samples

From: An unsupervised approach towards promptable porosity segmentation In laser powder bed fusion by segment anything

Samples

Dice Similarity Score

 

K-means

K-medoids

Reference binary masks

No Prompts

Sample 3

0.62 Âħ 0.034

0.62 Âħ 0.036

0.64 Âħ 0.029

0.0019 Âħ 0.0006

Sample 4

0.44 Âħ 0.12

0.44 Âħ 0.13

0.49 Âħ 0.11

0.0299 Âħ 0.0081

Sample 5

0.88 Âħ 0.019

0.88 Âħ 0.015

0.88 Âħ 0.016

0.0022 Âħ 0.0021

Sample 6

0.70 Âħ 0.092

0.75 Âħ 0.042

0.79 Âħ 0.098

0.0002 Âħ 0.0002

  1. The standard deviations are calculated to show the spread of the Dice Similarity Scores, indicating the variability of the results obtained from different prompt generation techniques across all samples. The first two columns show the DSC scores achieved using centroids derived from different clustering techniques respectively. The next column represents the DSC scores obtained from prompts generated from the respective reference binary masks of the individual image. The final column displays the DSC scores obtained without using any prompts.
Back to article page