dpvo ---colmap

[li1u]

Hello, I am amazed by the effect of your article. But I have some questions to ask you.

First, when I use video data, the visualization result is good, but the stored result.ply file is not good.

Second, there are still some differences between the cameras.txt and other files in the result folder and the standard colmap data. I don’t know whether your code provides an interface for storing as standard colmap data.

[lahavlipson]

Unfortunately, I am not able to reproduce this issue. If you could point me to an example which reproduces the problem, that would be very helpful for me.

The --colmap flag saves a directory which can be opened using:
colmap gui -> select file, import model -> select path/to/dpvo-output -> select no to only visualize reconstruction

When I do this on the demo from the colab notebook, I get:

FYI: The --colmap output does not contain information about the 2D feature tracks, since I assumed most people just wanted the camera poses and maybe a sparse reconstruction. If this would be helpful for people, I could add it.

[li1u]

Yes, it really works for me, hope you can add it.

[ljjTYJR]

@lahavlipson By the way, I think there is a minor error when saving the camera intrinsic parameter here:
In the COLMAP doc, when saving the camera parameters should be

# Camera list with one line of data per camera:
#   CAMERA_ID, MODEL, WIDTH, HEIGHT, PARAMS[]
# Number of cameras: 3
1 SIMPLE_PINHOLE 3072 2304 2559.81 1536 1152
2 PINHOLE 3072 2304 2560.56 2560.56 1536 1152
3 SIMPLE_RADIAL 3072 2304 2559.69 1536 1152 -0.0218531

but in the code, it is saved as

DPVO/dpvo/plot_utils.py

Line 56 in c0c5a10

(colmap_dir / "cameras.txt").write_text(f"1 PINHOLE {H} {W} {fx} {fy} {cx} {cy}")

I think it should be

 (colmap_dir / "cameras.txt").write_text(f"1 PINHOLE {W} {H} {fx} {fy} {cx} {cy}") 

[lahavlipson]

@ljjTYJR Thanks for catching that. It's now fixed