Frames order is reversed once I dump annotations

[muratayoshio]

My actions before raising this issue

• [ X] Read/searched the docs
• [X ] Searched past issues

Expected Behaviour

The frame sequence in annotations should respect the order of the uploaded sequence in the framework.

Current Behaviour

The frames sequence in annotations starts from the last frame and change the label id of the objects

Possible Solution

Steps to Reproduce (for bugs)

Context

Your Environment

• Git hash commit (git log -1):
• Docker version docker version (e.g. Docker 17.0.05):
• Are you using Docker Swarm or Kubernetes?
• Operating System and version (e.g. Linux, Windows, MacOS):
• Code example or link to GitHub repo or gist to reproduce problem:
• Other diagnostic information / logs:
  Logs from `cvat` container

Next steps

You may join our Gitter channel for community support.

[nmanovic]

@kirill-sizov , could you please look?

[sizov-kirill]

@kirill-sizov , could you please look?

@nmanovic, Yes, sure.

[sizov-kirill]

Hi, @muratayoshio,
Could you please describe in more details your steps leading to the problem? What format are you working with?

[muratayoshio]

Thank you...
I upload and annotated a video sequence and annotated the different objects which it appears in the Interface according to the normal video sequence, with lower Ids related to lower frames numbers.
Once I dump the annotation in MOT format or even in YOLO format, the gt txt file starts for the last frame, and assign object label sequence according to this reverse order. This off course will affect the tracking training and performance evaluation.
Thank you for your help.

[azhavoro]

@kirill-sizov Hi, what is the status of this issue?

[sizov-kirill]

@azhavoro Hi, not started yet, I will proceed with this task, after such as I finish my current task.

[sizov-kirill]

Hi, @muratayoshio

I still have problem with reproducing your issue.
I tried to follow your steps, but in all cases the gt.txt file was sorted by frame number.
There is how I annotated my task:

1. Create task with few labels and one video.
2. Put some rectangle shapes and tracks in the interface according to the normal video sequence.
3. Save
4. Dump annotations in MOT 1.1 format.

Could you please tell me at what step my actions differ from yours, so that I can reproduce this problem?

Thanks for your reply in advance.

[sizov-kirill]

I will close this issue as there is not enough information to reproduce it.

[ayhokuyan]

Hey @muratayoshio,
I was having the same trouble with the frames I have uploaded using the API requests. I was using python to parse the annotations and can offer a workaround. If you are using xml.etree.ElementTree to parse, you can sort by image id as follows. It is safer to say that I am using CVAT for Images 1.1 format, this can help someone at some point🙂

root = ET.parse(filename).getroot()
root[:] = sorted(root, key=lambda child: (child.tag, child.get('name')))

[muratayoshio]

Thank you very much, it is quite helpful... I'll try to implement this workaround then... El mar, 21 sept 2021 a las 9:34, Ayhan Okuyan ***@***.***>) escribió:

…

Hey @muratayoshio <https://github.com/muratayoshio>, I was having the same trouble with the frames I have uploaded using the API requests. I was using python to parse the annotations and can offer a workaround. If you are using xml.etree.ElementTree to parse, you can sort by image id as follows. root = ET.parse(filename).getroot() root[:] = sorted(root, key=lambda child: (child.tag, child.get('name'))) — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3366 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AGEALTTYXD63JPGP2BBDW7TUDCJYLANCNFSM47I5GNWA> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

--

___________________________ Jorge Ernesto Espinosa Oviedo PhD. Msc. Computer Engineer Docente Investigador Politécnico Colombiano Jaime Isaza Cadavid Tel. 57 (4) 319 79 00 ext 492 Oficina P19-211 G

[Chicone]

In my case, the reverse problem occurs when I load a bunch of images by dragging them into the loading area. If I create a video with the images and drag the video instead, the order is correct. To solve the problem of miss-assigned labels, you can upload the dumped (wrong) annotations into the same task and export it again (reverse + reverse = right order)

[nmanovic]

@Chicone , could you please describe how to reproduce the issue in a separate issue? Looks like it is a critical bug.