Replies: 3 comments 7 replies
-
from goodspb/pdlib#56 (comment) by @matiasdelellis Hi @luzfcb
Mainly it's great that you have an Nvidia GTX 1650 😬 , but the docker container is not compiled to use cuda. You should install CUDA on your host, and build the container to make use of it. 😉 On the other hand, you can reduce the size of temporary images on settings. By default this container is in 4Mpx, and with 2Mps you will get it will be 50% faster and you will get practically the same result. I left 4mpx as the default because the larger the images, the more likely you are to find all your faces, but in practice, you don't get more than 10% new faces. Finally no matter how long the process takes, it is designed to do the job progressively. We try not to take shortcuts but do better. p.s: Not even Google gives you the results so quickly when you upload many images... 😅 |
Beta Was this translation helpful? Give feedback.
-
I think one idea would be parallelization of the analyzation? Currrently it looks like only one image is analyzed at a time... WDYT @matiasdelellis ? |
Beta Was this translation helpful? Give feedback.
-
To speed up processin, I am following this guide and I am running 4 background jobs on the machine with 4 CPU cores. Very frecuently they keep having problem resolving external model docker container. Why is that? Faces found: 0. Image will be skipped because of the following error: Could not resolve host: nextcloud-aio-facerecognition
|
Beta Was this translation helpful? Give feedback.
-
I started testing Nextcloud AIO v7.10.0 Beta. Is there any way to improve performance of facerecognition-external-model?
For me, it analyzes 1 image every 16 seconds.
My server runs on Ryzen 5600G, 32 GB of RAM, and Nvidia GTX 1650 GPU, so I suppose it's more than enough to have a very good performance.
Beta Was this translation helpful? Give feedback.
All reactions