How to decide the "in_feature_dim" of train_**.py?

[GeoSur]

if self.config.in_features_dim == 1: pass elif self.config.in_features_dim == 4: stacked_features = np.hstack((stacked_features, features[:, :3])) elif self.config.in_features_dim == 5: stacked_features = np.hstack((stacked_features, features)) elif self.config.in_features_dim == 7: stacked_features = np.hstack((stacked_features, stacked_points, features[:, :3]))

As this section described, we could select 1,4,5 and 7 as the feature dim, but 1 and 4 represents without and with RGB respectively, but how to decide 5 and 7?

[LucaRom]

You can define the in_features_dim in the config class, for example S3DIS uses 5 features and it's defined on line 146 :

KPConv-PyTorch/train_S3DIS.py

Lines 144 to 146 in 435d117

	# Choice of input features
	first_features_dim = 128
	in_features_dim = 5