Training on data exported from Metashape

[ariannaferraris]

Hi there,

Firstly, thank you for your work!

I was trying to train the model on COLMAP data extracted from Metashape, that I managed to obtain through this script https://github.com/agisoft-llc/metashape-scripts/blob/master/src/export_for_gaussian_splatting.py

The folder structure I get after exporting my data from Metashape is the following:

• frame_000001

  • images
    • cam01_frame_000001
    • cam02_frame_000001
    • cam03_frame_000001
    • …
  • sparse
    • 0
      • cameras.bin
      • images.bin
      • points3D.bin

• frame_000002

  • images
    • cam01_frame_000002
    • cam02_frame_000002
    • cam03_frame_000002
    • …
  • sparse
    • 0
      • cameras.bin
      • images.bin
      • points3D.bin

• …

However, as soon as I try to train using the command python train.py -s {filedir} --port 6017 --expname "GS0003_Arianna_test" --configs arguments/hypernerf/default.py I get the following error:

Traceback (most recent call last):
File "train.py", line 428, in
training(lp.extract(args), hp.extract(args), op.extract(args), pp.extract(args), args.test_iterations, args.save_iterations, args.checkpoint_iterations, args.start_checkpoint, args.debug_from, args.expname)
File "train.py", line 302, in training
scene = Scene(dataset, gaussians, load_coarse=None)
File "...\4DGaussians\scene_init_.py", line 65, in init
assert False, "Could not recognize scene type!"
AssertionError: Could not recognize scene type!

I also tried without --configs arguments/hypernerf/default.py but I get the same error.

Could someone help me in understanding what’s wrong? If I try to train one single frame, I don’t have any problems and the result looks great, but now that I wanted to add the time dimension I cannot make it work. My setup consists of 42 static cameras and I was trying to train on 40 frames in total.

Thank you in advance!