Open AlexZou14 opened 3 years ago
i have the same question
Hi, The disparity maps contain some NAN values in the occlusion regions due to ZED cameras. Thus, it's correct when you read some NAN values in EXR files. In this paper, we remove the NAN and Inf in the Dataloader by setting these values to zeros: https://github.com/sczhou/DAVANet/blob/master/utils/data_loaders.py#L142 So that EPE loss will not be calculated in these regions.
We use your code to read the EXR file in the dataset and find that many of them have NAN values. Could you tell me how to right read EXR file or deal with these NAN values?