SJoJoK / 3DGStream

[CVPR 2024 Highlight] Official repository for the paper "3DGStream: On-the-fly Training of 3D Gaussians for Efficient Streaming of Photo-Realistic Free-Viewpoint Videos".
https://sjojok.github.io/3dgstream
MIT License
302 stars 18 forks source link

I have two doubts , #18

Closed AnudhinaDhanabalan closed 3 months ago

AnudhinaDhanabalan commented 3 months ago

i) I downloaded the FVV that you have provided in the 3dgstreamviewer and it is working ,but after i ran the extract _fvv.py i did not get the config.json file in the NTCs folder? ii) i have provided the model path containing the iteration 15000 ply file , but i my init_3dgs_ply file is not getting generated .can you please help me solve this error.

SJoJoK commented 3 months ago
  1. For the missing config.json, I suggest modifying the extract_fvv.py script to copy the file automatically. We'll address this in a future release.

  2. I need more details about the init_3dgs_ply file issue. Please specify the steps you followed and any error messages. Your current description doesn't give me a clear picture of what you are doing or what you expect to achieve.

AnudhinaDhanabalan commented 3 months ago

i provided the this as the path to the model path ""test/flame_steak_suite/flame_steak_init "" and ran the train_frames.py , by running the above i got the seperate ply files under iteration 150 and 250, but i did not get the init_3dgs_ply file anywhere

python train_frames.py --read_config --config_path test/flame_steak_suite/cfg_args.json -o temp/custom_op_stream -m test/flame_steak_suite/flame_steak_init -v dataset/flame_steak --image images --first_load_iteration 15000 --quiet

this is what i ran

SJoJoK commented 3 months ago

Oh I see, the so-called init_3dgs_ply is actually test/flame_steak_suite/flame_steak_init/point_cloud/iteration_15000/point_cloud.ply, you can copy it to you fvv directory and rename it.

AnudhinaDhanabalan commented 3 months ago

thank you so much for the clarification.

AnudhinaDhanabalan commented 3 months ago

can you please tell me what to do to get the config.json file ?

SJoJoK commented 3 months ago

It's the --ntc_conf_path when you run train_frames.py. In this case, it's configs/cache/cache_F_4.json. You can check test/flame_steak_suite/cfg_args.json to see what ntc_conf_path exactly is.

AnudhinaDhanabalan commented 3 months ago

thank youu so much !!!

SJoJoK commented 3 months ago

Glad to help:)