search

turandai / gaussian_surfels

Implementation of the SIGGRAPH 2024 conference paper "High-quality Surface Reconstruction using Gaussian Surfels".
537 stars 26 forks source link

Could not recognize scene type! #1

Closed AnudhinaDhanabalan closed 6 months ago

AnudhinaDhanabalan commented 6 months ago

I have followed all the steps properly with my own dataset. But during the training I am getting the error saying Traceback (most recent call last): File "train.py", line 300, in training(lp.extract(args), op.extract(args), pp.extract(args), args.test_iterations, args.save_iterations, args.checkpoint_iterations, args.start_checkpoint, args.debug_from) File "train.py", line 39, in training scene = Scene(dataset, gaussians, opt.camera_lr, shuffle=False, resolution_scales=[1, 2, 4]) File "/home/ghost/gaussian_surfels-main/scene/init.py", line 55, in init assert False, "Could not recognize scene type!" AssertionError: Could not recognize scene type!

-should I mention the path of the dataset anywhere in the code or how can I solve this error?

turandai commented 6 months ago

Hi, this assertion is raised here. The dataloader now can only accept 3 types of input foramt: COLMAP format with an 'sparse' subfolder, the NeRF Blender format with an transformstrain.json file, and IDR format with a 'cameras.npy' file. Please make sure to feed the right scene directory that contains the supported files. Hope this solves your problem.^^ I just see you mentioned you are using your own dataset. In case your dataset is in currently unsupported format, maybe you need to either transform your dataset into the supported formats, or write a custom dataloader to read your image and camera data.

AnudhinaDhanabalan commented 6 months ago

yeah but my dataset directory has the "sparse" subfolder , eventhough it is showing the same error

AnudhinaDhanabalan commented 6 months ago

solved the mistake ,thanks for replying !!