start colmap export #131
start colmap export #131
Conversation
gtsfm/utils/io.py
Outdated
| file_path = os.path.join(save_dir, "images.txt") | ||
| with open(file_path, "w") as f: | ||
| f.write("# Number of cameras: {}\n".format(gtsfm_data.number_images())) | ||
| f.write("# Image list with two lines of data per image:") | ||
| f.write("# IMAGE_ID, QW, QX, QY, QZ, TX, TY, TZ, CAMERA_ID, NAME") |
There was a problem hiding this comment.
Technically, camera ID and image ID can be different. How will we handle that? Maybe we need the camera-id image-id map in our pipeline too.
There was a problem hiding this comment.
i think that is a more conceptual change that will require some re-thinking. I added it to an issue #130
| # TODO: assign unique indices to all keypoints (2d points) | ||
| point2d_idx = 0 |
There was a problem hiding this comment.
These indices will be 0-n per image, right?
There was a problem hiding this comment.
I don't think so.
POINT2D_IDX defines the zero-based index of the keypoint in the images.txt file
There was a problem hiding this comment.
Actually, maybe you are right. It's a bit unclear.
There was a problem hiding this comment.
Maybe we should do a test run with COLMAP to check that
There was a problem hiding this comment.
@adisingh50 would you mind looking into this issue here about what the 2d keypoint/point index means?
|
Thanks for adding all this. Once this PR is merged into master, the point clouds we render will look much cooler now that we can render colors and camera poses. |
Dump
images.txt
cameras.txt
points3D.txt
per the COLMAP file specification