hi authors,
thanks for your great work. I have some training data questions, especially semanticKITTI
-
pre_train: I found you use the train data like 00 to 10 except 08. not the eval 08 and test data 11-21. Do I understand correctly? because it is reasonable
-
downstream train data. I found you use the datasets/percentiles_split.json to load the data. How is this json generated? is it randomly? As one simple way is use the python slicing, for example, if the train ration is 1%, and the skip ratio is 100, so man can easily use train_data[::100] to get it. Why you use a json file to load it?
hi authors,
thanks for your great work. I have some training data questions, especially semanticKITTI
pre_train: I found you use the train data like 00 to 10 except 08. not the eval 08 and test data 11-21. Do I understand correctly? because it is reasonable
downstream train data. I found you use the datasets/percentiles_split.json to load the data. How is this json generated? is it randomly? As one simple way is use the python slicing, for example, if the train ration is 1%, and the skip ratio is 100, so man can easily use train_data[::100] to get it. Why you use a json file to load it?