Closed ckLibra closed 3 years ago
Hi,
Our training/validation scene splits are randomly created while preparing the dataset (please see lines 248-279 here). As we noticed the splits given out by ScanNet authors at late stage in our work, we kept our random splits and I shared them for reproducibility of our experiments. Official ScanNetV2 test split is fixed for everyone: scene0707_00 to scene0806_00.
As far as I see, ESTDepth uses the splits produced by the ScanNet authors (compare here and here).
Hope this clarifies any confusion.
Thanks for your patient answer! It is very helpful.
Hi, Thanks for your great work! I have one question about the split of scannet. According to the train and validation.txt in your scannet-export folder, it seems a little different from https://github.com/xxlong0/ESTDepth/tree/main/data/scannet_split. I wonder which is the official split of the scannet.