Is it easy to support the training of images with both (H, W) and (W, H) sizes in the dataset?

[tb2-sy]

Dear authors,
Thank you for the great work. I have a question about the dataset training.
I tried to use the ActorHQ dataset for training on neus2, but the characteristics of the images in this dataset are images with both (H, W) and (W, H) placement, as shown in the picture above. Can your code support this? format data for training? Because I noticed that in the json example you provided, the size of the image in each frame is consistent.

[brianneoberson]

Hi, we're you able to run it on images with different sizes? I am also trying to test on a dataset that has images of different sizes, and I specify the height and width per frame in the transform.json, but I am getting errors.

[tb2-sy]

Hi, we're you able to run it on images with different sizes? I am also trying to test on a dataset that has images of different sizes, and I specify the height and width per frame in the transform.json, but I am getting errors.

I didn't solve this problem and gave up experimenting with this method.