About the dataset

[zhijieshen-bjtu]

Hi, I used the released code to train the models, but we got a significant better RMSE results (around 0.3). I check the max depth value in the provided dataset (the EXR files), and found that the max value is around 9.5 (metres?), is that right? It seems that the datasets has much smaller difference with those Indoor ones.

[lkku1]

Hi, what split sampling you take. I found they randomly split samples into training and validation datasets from the dataset with a ratio of 90% and 10% (each epoch or constant sampling).