Open suphamster opened 2 years ago
this made me laugh so hard
now this is some real "dickface" lol
So hilarious lmaooooo
now i'm trying if i can reproduce the bug
what model are u using for mosaic position detection? i'm using mosaic_position.pth
if there's bug it has to be problem with mosaic position model, maybe try to train a different model? (that may be hard)
Sadly I do not think there's a way to differentiate face and genital mosaic. @HypoX64 do you have any ideas?
hello, can you do me a favour? i'm running on NVIDIA GeForce GTX 1660 Ti (CUDA installed) and i'm working on a 3840x3210 video, but only have a speed of 0.05x (every 1 second in the video takes 20 seconds to process), and i'm not sure if this problem is specific to me, so can you run a test and hopefully calculate the speed?
hello, can you do me a favour? i'm running on NVIDIA GeForce GTX 1660 Ti (CUDA installed) and i'm working on a 3840x3210 video, but only have a speed of 0.05x (every 1 second in the video takes 20 seconds to process), and i'm not sure if this problem is specific to me, so can you run a test and hopefully calculate the speed?
Hi, I have only 2k JAV so far didn't saw any slow speeds while demosaic. But if u cut a 1 min fragment I could try... Are u using 11.0 cuda and GUI version or u run from source code?
i'm on ubuntu 21.10 using cuda 11.6. does newer version worsen performance? btw i'm running from source code
i'm on ubuntu 21.10 using cuda 11.6. does newer version worsen performance? btw i'm running from source code
I dunno but GUI+GPU version works strictly with 11.0. Also I heard that RTX 3000 Series needs CUDA 11.1 or newer (at least for JavPlayer v.1.09a + TG-Plus). But I think if u run from sources then it should be OK but latest Torch+Torchvision CUDA version is only v11.3 as far as I know according Pythorch.org (pip3 install torch==1.10.2+cu113 torchvision==0.11.3+cu113 torchaudio==0.10.2+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html) but I'm noob at Python so don't know for sure.
It may looks funny to see dickface ppl but maybe there is possible to auto detect faces and skip them or auto use faces models for it. Even cheap phones can detect faces so I think it's not difficult to imlement this (at least I hope so).