Urban area - Mix of buildings and vegetation

medium

#1

Dataset URL (Dropbox, Google Drive, GitHub…): https://drive.google.com/open?id=1Z-eG7pcHiBvGyODCcVoaz-H-mS5yhm1M
Number of images: 174
Size in MB: 1.6 GB
EXIF GPS (yes/no): yes
GCP (yes/no): no
RTK (yes/no):

Comment/Description: urban area. Mix of buildings and vegetation. Posting this data for testing…

It was taken using Pix4D capture app. I ran it in ODM : -
Left - Pix4D, Right - ODM
Nellore_2 Wall_of_building

Trying to figure out how to get sharp edges for buildings in ODM…The rest quality of point cloud is awesome!
Thanks for ODM community.


#2

Try with updated master, if you’re running on the command line. You should see better buildings with the latest master.


#3

Yup, try:

docker pull opendronemap/opendronemap

And then re-run the dataset with the latest code changes from the dem2mesh branch. https://github.com/OpenDroneMap/OpenDroneMap/pull/904

You can also try to tweak the --texturing-nadir-weight value (increase it to 32).


#4

The above picture was comparison of point cloud outputs, I am trying to get the boundaries of the buildings more clearer and sharper.

And I think --texturing-nadir-weight value is used at ODMMvsTexcell stage…so I think it will not be much use to improve the point cloud.
P.s : correct me if I am wrong. Thank you. :slight_smile:

Also tried with latest commit, but the point cloud looked same…


I havent checked improvements in orthophoto yet…will try it.


#5

My hunch is the lack of sharpness is a function of noise in the camera positions from OpenSfM. I know there’s work to improve the quality of the SfM model upstream in OpenSfM, but I haven’t yet heard the timing on that work. We likely will see improvements at this time.