Agisoft Metashape

Agisoft Metashape => Bug Reports => Topic started by: ParkerDevine on February 26, 2021, 04:21:08 AM

Title: 1.7.0: matchPhotos() and alignCameras() not producing any tie points
Post by: ParkerDevine on February 26, 2021, 04:21:08 AM
Hello, I am currently running into a major issue in v1.7.0 where tie points / key points are not being generated through the API.

I'm testing on a well-captured, 37 image project. Here is the code I am running.

import Metashape
doc = Metashape.app.document
chunk = doc.chunk

chunk.matchPhotos() #Tried with multiple parameter sets with no success
chunk.alignCameras() #Tried with multiple parameter sets with no success

All cameras and GPS data have been successfully uploaded to the project, so I didn't include the set up for that in the code above.

The problem is that while matchPhotos() and alignCameras() run without error, they do not generate any tie points. This is despite the fact that when running alignment through the GUI, >20,000 tie points are generated.

Is this a known bug for 1.7.0? If not, is there an extra step needed to generate sparse clouds in 1.7.0 that didn't exist in previous versions? I can't find any documentation about this issue in either the api manual or the forum
Title: Re: 1.7.0: matchPhotos() and alignCameras() not producing any tie points
Post by: Alexey Pasumansky on February 26, 2021, 10:57:55 AM
Hello ParkerDevine,

If the issue persists in 1.7.1 version, please send the related logs for both operations.
Title: Re: 1.7.0: matchPhotos() and alignCameras() not producing any tie points
Post by: Yoann Courtois on April 15, 2021, 03:48:48 PM
Hello Alexey,

We are facing the same issue in 1.7.1 release.
Here are some logs that might help you.

Regards
Title: Re: 1.7.0: matchPhotos() and alignCameras() not producing any tie points
Post by: Alexey Pasumansky on April 15, 2021, 08:16:47 PM
Hello Yoann,

Do you observe it for any project?
Title: Re: 1.7.0: matchPhotos() and alignCameras() not producing any tie points
Post by: Yoann Courtois on April 16, 2021, 11:31:01 AM
Yes of course, it happens on all our tests and with datasets we already processed without any problem in precious release
Title: Re: 1.7.0: matchPhotos() and alignCameras() not producing any tie points
Post by: Alexey Pasumansky on April 16, 2021, 02:40:28 PM
Hello Yoann,

I have used the script commands from the log that you have provided on some random projects in 1.7.2 version and every time have got proper tie point cloud:

Code: [Select]
def build_key_points(keypoint_limit, tiepoint_limit):
    Metashape.app.document.chunk.matchPhotos(downscale=1,  # HighAccuracy
                         generic_preselection=True,
                         reference_preselection=False,
                         keypoint_limit=keypoint_limit,
                         tiepoint_limit=tiepoint_limit,
                         keep_keypoints=False,
                         mask_tiepoints=False)
build_key_points(60000, 6000)

def build_tie_points():
    Metashape.app.document.chunk.alignCameras(adaptive_fitting=True)
build_tie_points()

Can you please share minimal set of images and the related problematic project saved in PSZ format after using these commands to support@agisoft.com?
Title: Re: 1.7.0: matchPhotos() and alignCameras() not producing any tie points
Post by: Yoann Courtois on April 16, 2021, 03:32:03 PM
I managed to get a project with proper point cloud on my computer, but with the same script commands a colleague of mine cannot reach the same result.

We will figure out where does the problem come from. It looks everything works well from 1.7.1 release.
Title: Re: 1.7.0: matchPhotos() and alignCameras() not producing any tie points
Post by: Alexey Pasumansky on April 16, 2021, 03:37:14 PM
Hello Yoann,

If the problem can be reproduced on any machine on your side, please provide the sample data and saved project to us for internal investigation.

Also please check, if the same version of Metashape is used in both "good" and "bad" cases and the conditions are the same: script is applied to the images in the newly added chunk (no existing tie point cloud or camera alignment).