search

fangchangma / sparse-to-dense.pytorch

ICRA 2018 "Sparse-to-Dense: Depth Prediction from Sparse Depth Samples and a Single Image" (PyTorch Implementation)
445 stars 101 forks source link

No rgb image normalization during pre-process #31

Closed Ellinier closed 4 years ago

Ellinier commented 4 years ago

Are rgb images normalized before input into the model for training? I don't find it and it seem that rgb images are only divided by 255 and transformed to tensor. If not, why depth estimation task doesn't need the normalization process, subtract the mean and then divide by the standard deviation, which is a routine process for other CV tasks such as semantic segmentation and object detection?

fangchangma commented 4 years ago

In our past experiments, the normalization of sparse depth does not make a significant difference in the results, and thus we didn't keep it.