Hi mates, i have a problem here. I just bought a RTX 3060 Ti from Maxsun, and it dont work with DeepFace.
When i put the mouse over the "Exclude GPU 0" in Global options, it seems to see the RTX.
I already uninstall all the versions of Conda and Python, and delete the Deepface folder, to make an Fresh install.
The program says at prompt:
"Setting Faceswap backend to NVIDIA"
But it still using 100% of the CPU, while the card and the VRAM dont change the usage.
Code: Select all
Loading...
Setting Faceswap backend to NVIDIA
05/11/2022 12:59:15 INFO Log level set to: INFO
05/11/2022 12:59:21 INFO Output Directory: E:\DeepFace\ProjDIY\FACE A
05/11/2022 12:59:25 INFO Loading Detect from S3Fd plugin...
05/11/2022 12:59:25 INFO Loading Align from Fan plugin...
05/11/2022 12:59:25 INFO Loading Mask from Components plugin...
05/11/2022 12:59:25 INFO Loading Mask from Extended plugin...
05/11/2022 12:59:25 INFO Loading Mask from Bisenet_Fp plugin...
05/11/2022 12:59:25 INFO Loading Mask from Vgg_Obstructed plugin...
05/11/2022 12:59:25 INFO Loading Mask from Unet_Dfl plugin...
05/11/2022 12:59:25 INFO Reset batch sizes due to available VRAM: Detect: 1, Align: 1, Mask: 1, Mask: 1, Mask: 1
05/11/2022 12:59:25 INFO Starting, this may take a while...
05/11/2022 12:59:25 INFO Initializing S3FD (Detect)...
05/11/2022 12:59:26 INFO Initialized S3FD (Detect) with batchsize of 1
05/11/2022 12:59:26 INFO Initializing FAN (Align)...
05/11/2022 12:59:45 INFO Initialized FAN (Align) with batchsize of 1
05/11/2022 12:59:45 INFO Initializing Components (Mask)...
05/11/2022 12:59:45 INFO Initialized Components (Mask) with batchsize of 1
05/11/2022 12:59:45 INFO Initializing Extended (Mask)...
05/11/2022 12:59:45 INFO Initialized Extended (Mask) with batchsize of 1
05/11/2022 12:59:45 INFO Initializing BiSeNet - Face Parsing (Mask)...
05/11/2022 12:59:52 INFO Initialized BiSeNet - Face Parsing (Mask) with batchsize of 1
05/11/2022 12:59:52 INFO Initializing VGG Obstructed (Mask)...
05/11/2022 13:00:03 INFO Initialized VGG Obstructed (Mask) with batchsize of 1
05/11/2022 13:00:03 INFO Initializing U-Net (Mask)...
05/11/2022 13:00:08 INFO Initialized U-Net (Mask) with batchsize of 1
The performance still at 6.3s/it. It would took a lot to finish just the extraction.
Any tips or anything that i could give to u to know it better?