Input format
Sapium59 opened this issue · comments
You mentioned that "The code takes a standard COLMAP workspace format as input". Does it mean I should run COLMAP first, before your script can generate personalized input for NeuconW?
I followed the COLMAP tutorial to produce a custom dataset. It has same structure as in 'Quickstart' chapter of tutorial. However, your preprocess shell script cannot work well with that, due to some path not matching correctly.
I also looked at LM dataset as file structure reference. It seems so different from colmap output, probably because it has been preprocessed. Would you like to offer a more detailed template, like a file path tree, for custom input?
update: when trying manually make custom trainjng data, find bad support in hard-code style
When work manually, step 1 and 2 go well. In step 3 generate cache, the 'data_generation.sh' script calls 'prepare_data_cache.py' in line 24 with 'dataset_name' kwarg explicitly defined as 'phototourism'. This causes a bug later when assigning 'depth_percentage', as my custom dataset has a different name and therefore not in your hard-coded dictionary in init method.
I am trying to get through your code and work out a solution.
update: worked out.
I changed downsample in config yaml,
and paths became matching well.
@zhu-yuefeng
Hello, may I ask you the way you solved the hard-code problem?
@zhu-yuefeng Hello, may I ask you the way you solved the hard-code problem?
I am truly sorry that I can hardly help.
After failing to get satisfying result on my own scene, the complete workspaced was removed. I can hardly recall any detail, either.
It is still recommended that, you place your run information (file structure, ...) and error infotmation (traceback, unit test, ...) here, and someone might join the discussion.
@zhu-yuefeng Hello, may I ask you the way you solved the hard-code problem?
I am truly sorry that I can hardly help. After failing to get satisfying result on my own scene, the complete workspaced was removed. I can hardly recall any detail, either. It is still recommended that, you place your run information (file structure, ...) and error infotmation (traceback, unit test, ...) here, and someone might join the discussion.
Thank you for a kind reply. I'm trying by myself and hopefully share my results if it goes well. Have a good day!