Error- "Our method cannot handle videos with large change of facial size!!"

[GayatriVadaparty]

Hey, Firstly I would like to thank for your work - DINet. I have set this up in an instance, created a virtual environment for this. I tried for example videos in asserts folder, and this is working great! I wanted to try for custom videos, which require landmarks csv file. For this, I've installed new version of openface. In that there's only option for webcam and I couldn't find a way to upload custom video and get landmarks csv. Then, I tried other scripts for getting landmarks to a video, when I inserted this file and ran the command it says "Our method cannot handle videos with large change of facial size!!"

Then I tried the openface 2.0.3 version and set up required settings for record, recording settings, openface setting, view, face detector, landmark detector as given in the repo. There I found three option for landmark detector : CLM, CLNF,CE-CLM. When I tried with CE-CLM, it says "Couldn't open the landmark detector module". When I tried with the other two, I got "video frames are misaligned with detected landmarks" How to fix this?

[Inferencer]

ce-clm is correct but your installation of openface is missing some files https://github.com/TadasBaltrusaitis/OpenFace/wiki/Windows-Installation#model-download

For CE-CLM landmark detector to work you need to download additional model files, this can be done by executing the download_models.ps1 PowerShell script. For more details on model download see - https://github.com/TadasBaltrusaitis/OpenFace/wiki/Model-download

[ingale726]

The same error, I downloaded four files, put in openface, can normal reasoning, but put the generated csv file into dinet reasoning error cen_patches_0.25_of.dat cen_patches_0.35_of.dat cen_patches_0.50_of.dat cen_patches_1.00_of.dat Openface versions 2.2.0,2.0.0, and 1.0.0 were tested,None of the generated csv files are available

the error:Our method cannot handle videos with large change of facial size!!

Which version should openface use? Or how should the generated csv file be post-processed?

[ingale726]

@GayatriVadaparty @Inferencer @MRzzm

[GayatriVadaparty]

@ingale726 Hey, I have got the same issue later when I tried.

[Inferencer]

The same error, I downloaded four files, put in openface, can normal reasoning, but put the generated csv file into dinet reasoning error cen_patches_0.25_of.dat cen_patches_0.35_of.dat cen_patches_0.50_of.dat cen_patches_1.00_of.dat Openface versions 2.2.0,2.0.0, and 1.0.0 were tested,None of the generated csv files are available

the error:Our method cannot handle videos with large change of facial size!!

Which version should openface use? Or how should the generated csv file be post-processed?

Lol that profile pic there are settings you need to select as mentioned in the repo when using openface anything not mentioned should be turned off/de-selected, no post processing needed on the csv I'm using version 1.0.0 If you continue to have issues you can let me know or add me on Discord: Inferencer and we can try some tests to figure out the cause

its also worth mentioning to set an output location for the csv too

[ingale726]

ok,I found the right setup

[Inferencer]

ok,I found the right setup

remove record au's and 3d landmarks & model parameters, should only have record 2d landmarks and record tracked images/videos selected

[ingale726]

ok,I found the right setup

remove record au's and 3d landmarks & model parameters, should only have record 2d landmarks and record tracked images/videos selected

OK