Question about the metrics of Table 4

Gavinwxy / DGCL

[CVPR 2023] Hunting Sparsity: Density-Guided Contrastive Learning for Semi-Supervised Semantic Segmentation

Apache License 2.0

23 stars 1 forks source link

Question about the metrics of Table 4 #9

Closed Hugo-cell111 closed 10 months ago

Hugo-cell111 commented 12 months ago

Hi! Could you provide the code of the metrics(Silhouette, Calinski-Harbasz and Davies-Boulding) shown in Table 4 in your paper? Thanks a lot!

Gavinwxy commented 10 months ago

Hi, thanks for your interest. The evaluation codes are unfortunately lost during code cleanning. They are implemented based on sklearn library, for example, the silhouette score and Calinski and Harabasz score. Before evaluation, I collected pixel-wise features from all categories and balanced their quantity. All features are evaluated with their pseudo labels.