[question] use colmap2nerf.py to extract 94 imgs on my own data. Only 2 imgs are logged in transforms.json. Why?

[zydmtaichi]

I run script to convert my video to nerf dataset. the program extract 94 imgs but only 2 of these are written to transforms.json. pls explain what problem cause this?

[casualdays]

I run script to convert my video to nerf dataset. the program extract 94 imgs but only 2 of these are written to transforms.json. pls explain what problem cause this?

Maybe you should improve the video (or image) quality you shoot (use 4k 60FPS), move slowly around the object when shooting, and preferably use an anti-shake device.

[TuXun1999]

I had a similar experience. My solution was to change the colmap camera model. By default, it's OPENCV. I changed it into PINHOLE, and successfully extracted out about 150 images from the video. To do this, check colmap2nerf.py

[zydmtaichi]

I had a similar experience. My solution was to change the colmap camera model. By default, it's OPENCV. I changed it into PINHOLE, and successfully extracted out about 150 images from the video. To do this, check colmap2nerf.py

yes, I get 92 extracted logs of 94 imgs in PINHOLE mode while only 2 extracted under OPENCV mode