Questions about certain code in fusion.py

[HuGhcat-code]

Dear author:
Thank you for sharing such a great piece of open-source work!
In fusion.py,I have questions about this part of the code：

features_mapping = features[:, mapping[:, 1], mapping[:, 2]]
features_mapping = features_mapping.permute(1, 0).cuda()
mask_k = mask != 0
#用于记录每个点在不同视图中被观察到的次数
gaussians._times[mask_k] += 1
#更新高斯模型的语义特征
gaussians._features_semantic[mask_k] += features_mapping[mask_k]

I don't understand what features_mapping and gaussians._times mean here
Also, why is the Gaussian semantic information from different views accumulated and then averaged?

[HuGhcat-code]

Hi,I can roughly judge that features_mapping refers to the features of the visible Gaussians in the current view, but I don't understand why the order of dimension 0 and dimension 1 of this tensor needs to be swapped using codefeatures_mapping = features_mapping.permute(1, 0).cuda()