Search found 22 matches

by police.bike
Thu Mar 04, 2021 3:07 pm
Forum: Extract Support
Topic: Use XSeg for masking
Replies: 1
Views: 9673

Use XSeg for masking

Is there a way to use XSeg with FaceSwap for masking ? I know XSeg is painful but it is so accurate esp with obstructions on face. Is there any way where I would be able to plug in a custom masker into faceswap. I dont hesitate coding for that. Any suggestions here are welcome. Existing vgg and unet...
by police.bike
Tue Feb 23, 2021 5:43 am
Forum: Cloud Discussion
Topic: Why the faceswap2.0 need python 3.7,on Google colab we only have python 3.6
Replies: 6
Views: 17608

Re: Why the faceswap2.0 need python 3.7,on Google colab we only have python 3.6

Update - Feb 20, 2021 I was able to get both python 3.7 & 3.8 working on google collab and train using latest versions. Pretty amazing ones ! Just use this as block 1 before running any steps within the Collab notebook !sudo apt install python3.7 !update-alternatives --install /usr/local/bin/pyt...
by police.bike
Tue Feb 23, 2021 5:43 am
Forum: Cloud Discussion
Topic: [Outdated] Training Using Google Colab
Replies: 66
Views: 138029

Re: [Resource] Training Using Google Colab

Update - Feb 20, 2021 I was able to get both python 3.7 & 3.8 working on google collab and train using latest versions. Pretty amazing ones ! Just use this as block 1 before running any steps within the Collab notebook !sudo apt install python3.7 !update-alternatives --install /usr/local/bin/pyt...
by police.bike
Mon Nov 23, 2020 6:37 pm
Forum: Cloud Discussion
Topic: Latest release faceswap with Vast.ai server setup - training
Replies: 0
Views: 25818

Latest release faceswap with Vast.ai server setup - training

If you wonder how to train using vast.ai or any cloud servers with latest faceswap or if you run across issues with r1.0 edition on cloud servers, then you are like me. After much of research and reading, below is what got me a successful cloud training server. The configuration is for Ubuntu server...
by police.bike
Wed Nov 18, 2020 1:04 am
Forum: Cloud Discussion
Topic: [Guide] Vast.ai setup script
Replies: 14
Views: 42021

Re: [Guide] Vast.ai setup script

This procedure allow use gui? Im looking for alternative to Google Cloud because i want try different gpu and google allow only Tesla You can use gnome panel and a bunch of utilities using a initial shell connection to your vm. Install the terminal, xfce and other goodies that you need for desktop ...
by police.bike
Wed Nov 18, 2020 1:04 am
Forum: Cloud Discussion
Topic: [Guide] Vast.ai setup script
Replies: 14
Views: 42021

Re: [Guide] Vast.ai setup script

command line will be faster and better. But yes, you can have it traditional way using a desktop config onto your remote

by police.bike
Sat Aug 29, 2020 7:58 pm
Forum: Cloud Discussion
Topic: [Guide] Vast.ai setup script
Replies: 14
Views: 42021

Re: [Guide] Vast.ai setup script

Can you add the below line to the guide :

apt-get install -y libgl1-mesa-dev

must be done if faceswap errors out. Its something to do with libGL 1.0
This fixed my problems and was able to use vast server on 1.15.2-gpu-py3-jupyter

by police.bike
Sat Aug 29, 2020 12:38 am
Forum: Cloud Discussion
Topic: [Guide] Vast.ai setup script
Replies: 14
Views: 42021

Re: [Guide] Vast.ai setup script

no luck even after switching to r1.0
I tried with 1.15.2-gpu-python3-jupyter and same image with latest

by police.bike
Fri Aug 28, 2020 5:47 pm
Forum: Cloud Discussion
Topic: [Guide] Vast.ai setup script
Replies: 14
Views: 42021

Re: [Guide] Vast.ai setup script

no luck. Seems faceswap got updated recently and the clone creates this error. Any help is much appreciated

by police.bike
Fri Aug 28, 2020 1:19 pm
Forum: Cloud Discussion
Topic: [Guide] Vast.ai setup script
Replies: 14
Views: 42021

Re: [Guide] Vast.ai setup script

I think this isn't working anymore. While installing Git, it forces a apt update and otherwise git won't install. Once updated, a faceswap clone and nvidia installation is done. But when I try to set backend by running python faceswap.py I get an error saying python 3.7 is required. I installed it u...
by police.bike
Sat Jul 04, 2020 2:52 am
Forum: Training Support
Topic: application wont run on GPU, low EGs
Replies: 6
Views: 3545

Re: application wont run on GPU, low EGs

My 2 cents... I had a similar lower end graphics and I stayed with Original Model. By adjusting coverage % to > 82% and opting ssim in loss function for training, i was able to get very good result.

by police.bike
Fri Jul 03, 2020 11:34 pm
Forum: Training Discussion
Topic: how does multi GPU training work
Replies: 5
Views: 3455

how does multi GPU training work

I am using vast to render multiple GPU processing. I am curious how it works. Say if I choose a 2 GPU machine and set bs to 64. Does it branch 32 on one and another 32 on the other ? Or does bs indicate a 64 on each GPU. Curious to know how it works. It was surprising that a 2X 11 G NVIDIA machines ...
by police.bike
Fri Jul 03, 2020 4:47 pm
Forum: Cloud Discussion
Topic: [Guide] Vast.ai setup script
Replies: 14
Views: 42021

Re: [Guide] Vast.ai setup script

Had some issues with my home computer so used vast.ai to rent instances. A GTX 1080-TI runs about 0.1$/hour. When creating instances use the tensorflow/tensorflow:1.15.2-gpu image Here's the script I run to setup the enviroment, put it in a .sh file. Open a terminal and do a "chmod +x" on...
by police.bike
Fri Jul 03, 2020 4:01 pm
Forum: Training Discussion
Topic: Can Training set be consist of different persons' faces?
Replies: 2
Views: 1888

Re: Can Training set be consist of different persons' faces?

I followed the extract guide and put the different persons' faces into one files , then merged the alignments. Can them be trained at the same time by the model? For example, if I want to swap B into video1(contains A1 faces) and video2(contains A2 faces), can I train model for A1 and A2 simultaneo...
by police.bike
Fri Jul 03, 2020 3:56 pm
Forum: Training Discussion
Topic: Changing wrap to landmarks in middle of training - needs a new model creation ?
Replies: 1
Views: 2476

Changing wrap to landmarks in middle of training - needs a new model creation ?

Does changing 'Wrap to landmarks' option from false to true in middle of a training with 65k iterations, needs creating a new model? Or does it support toggling on an existing training model ... sorry to bug with so many questions in recent days. But I got to tell you, I love this much more than DFL...
by police.bike
Thu Jul 02, 2020 1:16 am
Forum: Extract Discussion
Topic: Does Cv2-dnn extractor means inferior alignments ?
Replies: 2
Views: 2408

Re: Does Cv2-dnn extractor means inferior alignments ?

------------- Did a bit of research myself by extracting and converting with all the 3 detectors. HERE are my findings: Cv2-Dnn - Gave me lesser number of faces. As admin mentioned, we need to run multiple passes. However, sequence of images is not guaranteed. It still missed a lot of faces. On conv...
by police.bike
Wed Jul 01, 2020 10:42 pm
Forum: Extract Support
Topic: Face filter or NFilter files downloaded/decompressed but nothing happens after that
Replies: 1
Views: 1663

Face filter or NFilter files downloaded/decompressed but nothing happens after that

During extraction, when I add a Filter or NFilter under FaceProcessing, I notice that the zip file of about 580Mb is getting downloaded. It decompresses and then nothing happens. I waited for a while and then removed that option and proceeded. I checked my firewall and permissions and it looks fine....
by police.bike
Wed Jul 01, 2020 9:48 pm
Forum: Extract Discussion
Topic: Does Cv2-dnn extractor means inferior alignments ?
Replies: 2
Views: 2408

Does Cv2-dnn extractor means inferior alignments ?

While extract , it is mentioned that cv2-dnn is CPU only while Sf3d is recommended. Does it mean we should not use cv2-dnn at all ? It seems fast. So, Does it mean the quality or alignments will be of inferior quality? Is the only difference between these detectors/aligners are about finding a face ...
by police.bike
Wed Jul 01, 2020 9:45 pm
Forum: Training Discussion
Topic: Some swaps still looks like A->A instead of A->B
Replies: 2
Views: 2018

Re: Some swaps still looks like A->A instead of A->B

sure. will post. I switched to Original with a reduced coverage % and mae loss when I noticed this issue. Will try to re-create the scenario and update. Thanks for your reply.

by police.bike
Tue Jun 30, 2020 4:12 pm
Forum: Training Discussion
Topic: Some swaps still looks like A->A instead of A->B
Replies: 2
Views: 2018

Some swaps still looks like A->A instead of A->B

Sometimes I had a strange training output with below settings: Original model -> 87.5 coverage -> Batch 16/32/64 -> smooth/ssim -> Extended mask -> default kernel blur on masks The preview after 50k iterations of batch size 64 still looks like A->A rather than A->B. Esp with ssim loss, I see this of...