search

InternLandMark / LandMark

Other
452 stars 38 forks source link

How to convert the contextcapture result to LandMark dataset format #5

Closed chufall closed 1 year ago

chufall commented 1 year ago

Hi, this is an amazing work!

I have > 1000 images taken by a drone and reconstructed by ContextCapture as same as what mentioned in your grid-nerf paper. So how to convert the result of ContextCapture into the LandMark dataset format ? Is there any code or pipeline to do it? Thanks a lot!

chu

kam1107 commented 1 year ago

Thanks for your interests! Please first export xml pose file (under ENU coordinate) from ContextCapture. Here's a reference code of converting the xml pose file to our pose format (.json): https://drive.google.com/file/d/1Hilc93M2G0rLatz093REilrT6Rub8a5c/view?usp=sharing You may want to change the way to iterate through photos/photo groups based on your own xml file.

chufall commented 1 year ago

Thanks a lot !!! I will try!

chufall commented 1 year ago

@kam1107 There're some bugs in your code, and I have fixed it. https://drive.google.com/file/d/1tFvti9y1fv7XflVI2Fyt6NEsJQYx6FIV/view?usp=sharing

Here's something need to be confirmed:

  1. The result of the code is a json file named drone_3.json, whose format is different with which need by CityDataset class (transforms_train.json and transforms_test.json) , should I convert it?

  2. When I used the tools/colmap2nerf.py to convert a colmap dataset, and trained it , the CityDataset class will return a error "KeyError: 'focal'" in line 113 of app/tools/dataloader/ray_utils.py I found there isn't the key "focal" in the transforms_train.json outputed by colmap2nerf.py.

Thanks a lot!

FlushingCat commented 1 year ago

Hi! about the "KeyError: 'focal'" problem, this thread may help you Issues: KeyError: 'focal' #1

chufall commented 1 year ago

@FlushingCat
I got it!

chufall commented 1 year ago

@kam1107 @FlushingCat I have successfully submited the traning job! Thanks a lot！

chufall commented 7 months ago

@kam1107 There're some bugs in your code, and I have fixed it. https://drive.google.com/file/d/1tFvti9y1fv7XflVI2Fyt6NEsJQYx6FIV/view?usp=sharing Here's something need to be confirmed:

  1. The result of the code is a json file named drone_3.json, whose format is different with which need by CityDataset class (transforms_train.json and transforms_test.json) , should I convert it?
  2. When I used the tools/colmap2nerf.py to convert a colmap dataset, and trained it , the CityDataset class will return a error "KeyError: 'focal'" in line 113 of app/tools/dataloader/ray_utils.py I found there isn't the key "focal" in the transforms_train.json outputed by colmap2nerf.py.

Thanks a lot!

@chufall Do you solve this problem?json format is different with which need by CityDataset class (transforms_train.json and transforms_test.json)" please help me out!!!

drone2nerf.zip

Hi, sorry for later! Try my code, and the command line is the following --drone_xml xxx\blockexchange.xml --output_dir xxxx\output