@ -231,12 +231,12 @@ Immich optionally uses machine learning for several features. However, it can be
### Can I lower CPU and RAM usage?
The initial backup is the most intensive due to the number of jobs running. The most CPU-intensive ones are transcoding and machine learning jobs (Tag Images, Smart Search, Recognize Faces), and to a lesser extent thumbnail generation. Here are some ways to lower their CPU usage:
The initial backup is the most intensive due to the number of jobs running. The most CPU-intensive ones are transcoding and machine learning jobs (Smart Search, Face Detection), and to a lesser extent thumbnail generation. Here are some ways to lower their CPU usage:
- Lower the job concurrency for these jobs to 1.
- Under Settings > Transcoding Settings > Threads, set the number of threads to a low number like 1 or 2.
- Under Settings > Machine Learning Settings > Facial Recognition > Model Name, you can change the facial recognition model to `buffalo_s` instead of `buffalo_l`. The former is a smaller and faster model, albeit not as good.
- You _must_ re-run the Recognize Faces job for all images after this for facial recognition on new images to work properly.
- You _must_ re-run the Face Detection job for all images after this for facial recognition on new images to work properly.
- If these changes are not enough, see [below](/docs/FAQ#how-can-i-disable-machine-learning) for how you can disable machine learning.
### Can I limit the amount of CPU and RAM usage?
@ -247,10 +247,10 @@ You can look at the [original docker docs](https://docs.docker.com/config/contai
### How an I boost machine learning speed?
:::note
This advice increases throughput, not latency. This is to say that it will make Smart Search jobs process more quickly, but it won't make searching faster.
This advice improves throughput, not latency. This is to say that it will make Smart Search jobs process more quickly, but it won't make searching faster.
:::
You can increase throughput by increasing the job concurrency for machine learning jobs (Smart Search, Recognize Faces). With higher concurrency, the host will work on more assets in parallel. You can do this by navigating to Administration > Settings > Job Settings and increasing concurrency as needed.
You can increase throughput by increasing the job concurrency for machine learning jobs (Smart Search, Face Detection). With higher concurrency, the host will work on more assets in parallel. You can do this by navigating to Administration > Settings > Job Settings and increasing concurrency as needed.
:::danger
On a normal machine, 2 or 3 concurrent jobs can probably max the CPU, so if you're not hitting those maximums with, say, 30 jobs.
subtitle: 'Run machine learning on assets to support smart search',
disabled: !$featureFlags.clipEncode,
},
[JobName.RecognizeFaces]: {
[JobName.FaceDetection]: {
icon: mdiFaceRecognition,
title: api.getJobName(JobName.RecognizeFaces),
subtitle: 'Run machine learning on assets to recognize faces',
handleCommand: handleFaceCommand,
title: api.getJobName(JobName.FaceDetection),
subtitle:
'Detect the faces in assets using machine learning. For videos, only the thumbnail is considered. "All" (re-)processes all assets. "Missing" queues assets that haven\'t been processed yet. Detected faces will be queued for Facial Recognition after Face Detection is complete, grouping them into existing or new people.',
handleCommand: handleConfirmCommand,
disabled: !$featureFlags.facialRecognition,
},
[JobName.FacialRecognition]: {
icon: mdiTagFaces,
title: api.getJobName(JobName.FacialRecognition),
subtitle:
'Group detected faces into people. This step runs after Face Detection is complete. "All" (re-)clusters all faces. "Missing" queues faces that don\'t have a person assigned.',
handleCommand: handleConfirmCommand,
disabled: !$featureFlags.facialRecognition,
},
[JobName.VideoConversion]: {
@ -131,11 +144,11 @@
}
</script>
{#iffaceConfirm}
{#if confirmJob}
<ConfirmDialogue
prompt="Are you sure you want to reprocess all faces? This will also clear named people."
desc="Models are listed in descending order of size. Larger models are slower and use more memory, but produce better results. Note that you must re-run the Recognize Faces job for all images upon changing a model."
desc="Models are listed in descending order of size. Larger models are slower and use more memory, but produce better results. Note that you must re-run the Face Detection job for all images upon changing a model."
desc="The minimum number of faces of a person that must be detected for them to appear in the People tab. Setting this to a value greater than 1 can prevent strangers or blurry faces that are not the main subject of the image from being displayed."
label="MIN RECOGNIZED FACES"
desc="The minimum number of recognized faces for a person to be created. Increasing this makes Facial Recognition more precise at the cost of increasing the chance that a face is not assigned to a person."
Mert
GitHub