Search found 1960 matches

by torzdf
Fri Sep 23, 2022 5:20 pm
Forum: Extract Support
Topic: Error Logged While Extracting - 32bit limitation?
Replies: 3
Views: 24

Re: Error Logged While Extracting - 32bit limitation?

Ah, ok. The res may well be why. I will artificially upscale some videos for testing. I doubt that this would have caused any issues with output, but I can't guarantee it. I will look at what returns from numpy when this warning is hit. However, as it is just a test on whether the face is bigger tha...
by torzdf
Fri Sep 23, 2022 2:17 pm
Forum: Extract Support
Topic: Error Logged While Extracting - 32bit limitation?
Replies: 3
Views: 24

Re: Error Logged While Extracting - 32bit limitation?

Ok, that's annoying. This is part of the new filters, of which I have run literally hundreds of thousands of frames through now and not hit this issue :/ If there is anyway you can provide me a source file + config which reproduces this I would be able to look for a fix. In the meantime you can disa...
by torzdf
Thu Sep 22, 2022 11:15 am
Forum: Training Support
Topic: Caught exception in thread: '_training'
Replies: 9
Views: 396

Re: Caught exception in thread: '_training'

I haven't forgotten about this, just need to find some time to troubleshoot

by torzdf
Thu Sep 22, 2022 11:14 am
Forum: Training Support
Topic: Weird Faceswap behaviour [ KeyError: 'faceswap' ]
Replies: 2
Views: 13

Re: Weird Faceswap behaviour [ KeyError: 'faceswap' ]

I'm glad, because I was struggling to understand what was causing the issue :wink:

by torzdf
Thu Sep 22, 2022 11:13 am
Forum: Extract Support
Topic: landmarks for profile image
Replies: 4
Views: 228

Re: landmarks for profile image

The landmarks for "off-frame" are placed where the net "thinks" they should be, so they will be positioned off-frame

by torzdf
Mon Sep 19, 2022 3:57 pm
Forum: Extract Support
Topic: landmarks for profile image
Replies: 4
Views: 228

Re: landmarks for profile image

They go flush with the "horizon" of the face iirc. Sometimes in extreme profiles they will be shown "through" the face, but generally those landmarks will be lined up on the edge of the visible face.

by torzdf
Sun Sep 18, 2022 5:32 pm
Forum: Installation Support
Topic: Windows Installer error: “Some applications failed to install.Process Aborted.Check Details.”
Replies: 5
Views: 842

Re: Windows Installer error: “Some applications failed to install.Process Aborted.Check Details.”

This is a connection error. Try the steps posted in the message directly prior to your one

by torzdf
Sun Sep 18, 2022 5:30 pm
Forum: Training Support
Topic: Caught exception in thread: '_training'
Replies: 9
Views: 396

Re: Caught exception in thread: '_training'

Ok, I figured that was unlikely. The issue appears to be the application of masks in the loss function. As this is not a widely reported issue, it is either specific to what you are doing, or it is due to the combination of some niche settings. Can you please try the following: Attempt training an o...
by torzdf
Sun Sep 18, 2022 5:26 pm
Forum: Extract Support
Topic: Help me fix "OMP: Error #15: Initializing libiomp5md.dll, but found libiomp5 already initialized."
Replies: 1
Views: 186

Re: Help me fix "OMP: Error #15: Initializing libiomp5md.dll, but found libiomp5 already initialized."

Yes, these seems to be a CPU on Windows only specific issue.

I have pushed an update to suppress this error. Please update to see if it solves your issue

by torzdf
Sun Sep 18, 2022 5:15 pm
Forum: Training Discussion
Topic: Dynamic algorithmic input source trimming while training
Replies: 5
Views: 322

Re: Dynamic algorithmic input source trimming while training

Do batches always contain the same group of face images? No they are random. After a certain number of epochs you should be able to predict, with a fair degree of certainty, which individual images impact batch scores the most. I'm making the mistake of brainstorming without intricate knowledge of ...
by torzdf
Sat Sep 17, 2022 11:33 am
Forum: Training Support
Topic: Caught exception in thread: '_training'
Replies: 9
Views: 396

Re: Caught exception in thread: '_training'

Ok, it's a long shot, but in the first instance, could you please uninstall the globally installed "Cuda 11.7" and see if that makes a difference.

by torzdf
Fri Sep 16, 2022 11:22 am
Forum: Extract Support
Topic: Status: Failed - extract.py. Return Code: 3221226505
Replies: 1
Views: 212

Re: Status: Failed - extract.py. Return Code: 3221226505

I suspect this is a memory error due to being a fairly small GPU.

Firstly, make sure your GPU drivers are up to date.
Secondly, try using the MTCNN detector and see if the problem persists.

by torzdf
Fri Sep 16, 2022 11:18 am
Forum: Introductions
Topic: Faceswap installation not working on my PC.
Replies: 7
Views: 375

Re: Faceswap installation not working on my PC.

Ok, I'm not sure I'm going to be able to solve this. As artisan says, you have posted that you are using 32bit Python, which is not supported. Our standard install path is to run the Faceswap installer from: https://forum.faceswap.dev/viewtopic.php?t=20. Manual installs are, by their nature, very ha...
by torzdf
Fri Sep 16, 2022 11:16 am
Forum: Training Support
Topic: Caught exception in thread: '_training'
Replies: 9
Views: 396

Re: Caught exception in thread: '_training'

Code: Select all

Crash report written to 'C:\Users\adama\faceswap\crash_report.2022.09.15.101147088901.log'. You MUST provide this file if seeking assistance
by torzdf
Fri Sep 16, 2022 11:14 am
Forum: Training Discussion
Topic: Yet, Another Loss Value Question 😁
Replies: 9
Views: 466

Re: Yet, Another Loss Value Question 😁

It shouldn't make a difference. The reason for the color weirdness is that most models are in BGR color order (openCVs color format) but some are in RGB order (most keras encoders), so we have to display the previews correctly (as I said, the model is fine, this just impact previews). Basically, in ...
by torzdf
Fri Sep 16, 2022 11:11 am
Forum: Extract Discussion
Topic: Matching extraction "size" configuration with selected model type
Replies: 6
Views: 370

Re: Matching extraction "size" configuration with selected model type

I highly doubt the image will be overly sharp. I wouldn't imagine anything detrimental would come from this approach, just that you wouldn't necessarily be using the model to its full capacity.

by torzdf
Fri Sep 16, 2022 11:10 am
Forum: Training Discussion
Topic: Dynamic algorithmic input source trimming while training
Replies: 5
Views: 322

Re: Dynamic algorithmic input source trimming while training

There are many potential experiments that could be run on data feeding. How useful they would or would not be is an open question. It wouldn't be possible to track score per image (unless validating on a batch size of 1) but scores could be kept at a batch level, and tracked over time, looking to re...
by torzdf
Thu Sep 15, 2022 12:48 pm
Forum: Introductions
Topic: Faceswap installation not working on my PC.
Replies: 7
Views: 375

Re: Faceswap installation not working on my PC.

Most likely you are not activating the faceswap environment.

It's hard to troubleshoot without knowing the python setup/install environments etc. that you have.

by torzdf
Wed Sep 14, 2022 11:13 am
Forum: Artificial Intelligence
Topic: Does Faceswap have GAN training like Deepfacelab?
Replies: 1
Views: 256

Re: Does Faceswap have GAN training like Deepfacelab?

Experiments with GANs have, so far, been unsatisfying, so there is no GAN in Faceswap.

You can also look into the thread on loss functions for other 'sharpness' techniques
viewtopic.php?t=2176

by torzdf
Wed Sep 14, 2022 11:10 am
Forum: Training Support
Topic: Only getting 10EG/s and GPU showing only 1% util in nvidia-smi
Replies: 1
Views: 219

Re: Only getting 10EG/s and GPU showing only 1% util in nvidia-smi

The speed is hard to say without knowing the model.

If that Nvidia-SMI screengrab is from when Faceswap is running, then it would suggest your GPU is not being used. Please post the output of (from GUI) help > output system info