The multi-view images and depth maps generated by training with your own data set are blank

[yuanjie0217]

I referred to the format of the nerf data set and used colmap to generate the pose file. The image is my own simulated SAR image with a size of 4032*3024, which is the same size as the image in the images file in the LLFF data set. Then I input my data set into the nerf variant model of Sparf(https://github.com/google-research/sparf) for training, but the output structure was all null values, and the loss value remained unchanged at 60. I tried the combination of my data set image + the pose file in the open source LLFF data set and the image in the open source LLFF data set + the pose file of my custom data set, and found that both results were output. Does this mean that my data set Problem with files generated via colmap?