search

kwea123 / nerf_pl

NeRF (Neural Radiance Fields) and NeRF in the Wild using pytorch-lightning
https://www.youtube.com/playlist?list=PLDV2CyUo4q-K02pNEyDr7DYpTQuka3mbV
MIT License
2.74k stars 483 forks source link

use colmap_colab run in silica data get a error #118

Closed Yes-Jumby closed 2 years ago

Yes-Jumby commented 2 years ago

dear auther, all is same to the demo like you.but get image

i note in original ipynb,here is the result: image IImages # 65 while it is 35 in my output I get 48 imgs from your google driver,maybe the data has difference?

Yes-Jumby commented 2 years ago

I check it ,in the video the img is from IMG_E4511 +; while in my download data set ,it is IMG_E4615.JPG to IMG_E4662JPG(in your release link:https://drive.google.com/drive/folders/1qdZD7UgUzQIc52-xQ83QIJbFsvnXEkX2); could you please share the original dataset?

teachyourselfcoding commented 2 years ago

I trained using the colmap output provided and it was successful. I tried to run colmap myself using the images from the google drive and I had the same error as you. Do @me if you managed to fix this issue!

kwea123 commented 2 years ago

Hi, indeed the data in the video is different with what is on the google drive now. I forgot the reason why I retake another dataset, but I have made sure that the data is consistent. The 48-image data is the correct one. Sometimes COLMAP fails to get the pose for all images (in your case only 35/48 was successful), that makes the imgs2poses.py fail. I recommend using my poses_bounds.npy directly on the google drive, and you can skip the colmap notebook.

teachyourselfcoding commented 2 years ago

I see, any idea why colmap cant find poses for all images?

teachyourselfcoding commented 2 years ago

colmap每次的結果會有隨機性，所以有時候會失敗。這個只能通過調整參數的方法來試誤。我colab上提供的沒有這個功能，你需要查詢一下相關命令，然後修改colab上的命令才可以

There is a randomness to the colmap output, which would require adjusting parameters of colmap, which is a feature not available on the colab notebook.

-from your reply elsewhere, thanks!