@sthropic seems like I was too slow and you solved it already! Your solution worked fine for me. Great job! For the sake of completeness: I used "photoprism faces reset --engine=pigo" or the environment variable "PHOTOPRISM_FACE_ENGINE=pigo".
Search Criteria
Package Details: photoprism 260601-1
Package Actions
| Git Clone URL: | https://aur.archlinux.org/photoprism.git (read-only, click to copy) |
|---|---|
| Package Base: | photoprism |
| Description: | AI-Powered Photos App for the Decentralized Web |
| Upstream URL: | https://github.com/photoprism/photoprism |
| Keywords: | ai cloud golang photography |
| Licenses: | AGPL-3.0-or-later |
| Submitter: | evine |
| Maintainer: | shtrophic |
| Last Packager: | shtrophic |
| Votes: | 3 |
| Popularity: | 0.005582 |
| First Submitted: | 2024-06-18 05:57 (UTC) |
| Last Updated: | 2026-06-05 15:23 (UTC) |
Dependencies (16)
- glib2 (glib2-gitAUR, glib2-patched-thumbnailerAUR)
- glibc (glibc-gitAUR, glibc-eacAUR, glibc-git-native-pgoAUR)
- libvips (libvips-gitAUR, libvips-notiffAUR)
- photoprism-onnxruntime (photoprism-onnxruntime-binAUR, photoprism-onnxruntime-cuda-binAUR)
- photoprism-tensorflow (photoprism-tensorflow-binAUR, photoprism-tensorflow-avx512-binAUR, photoprism-tensorflow-vnni-binAUR, photoprism-tensorflow-avx2-binAUR, photoprism-tensorflow-avx-binAUR, photoprism-tensorflow-ssse3-binAUR, photoprism-tensorflow-cuda-binAUR)
- findutils (findutils-gitAUR, findutils-selinuxAUR) (make)
- go (go-gitAUR, gcc-go-gitAUR, goup-rsAUR, go-binAUR, gcc-go-snapshotAUR, gcc-go) (make)
- npm (npm-corepackAUR, python-nodejs-wheelAUR) (make)
- darktable (darktable-gitAUR) (optional) – for RAW to JPEG conversion
- ffmpeg (ffmpeg-nvcodec-11-1-gitAUR, ffmpeg-ffplayoutAUR, ffmpeg-cudaAUR, ffmpeg-decklinkAUR, ffmpeg-amd-fullAUR, librempeg-gitAUR, ffmpeg-whisper-gitAUR, ffmpeg-gitAUR, ffmpeg-fullAUR, ffmpeg-full-llvmAUR, ffmpeg-libfdk_aacAUR, ffmpeg-obsAUR, ffmpeg-headlessAUR, ffmpeg-whisperAUR, ffmpeg-cuda-fullAUR, ffmpeg-full-gitAUR, ffmpeg-amd-full-gitAUR) (optional) – for video transcoding and thumbnail extraction
- imagemagick (imagemagick-full-gitAUR, imagemagick-gitAUR, imagemagick-essentials-gitAUR, imagemagick-fullAUR) (optional) – for image conversion
- libheif (libheif-gitAUR, libheif-highmemAUR, libheif-noglycinAUR) (optional) – for HEIC/HEIF image conversion
- libva (libva-gitAUR, intel-media-stack-legacy-binAUR, intel-media-stack-binAUR, libva-headlessAUR) (optional) – for video transcoding
- mariadb (mysql81AUR, mysql80AUR, mysql84AUR, mysqlAUR, mariadb-lts, percona-server) (optional) – for MySQL database driver
- perl-image-exiftool (optional) – for extracting metadata
- rawtherapee (rawtherapee-gitAUR) (optional) – for RAW to JPEG conversion
Required by (0)
Sources (15)
- 01-internal-config-config_go.patch
- defaults.yml
- https://dl.photoprism.app/onnx/models/scrfd_500m_bnkps_shape640x640.onnx
- https://dl.photoprism.app/tensorflow/facenet.zip
- https://dl.photoprism.app/tensorflow/nasnet.zip
- https://dl.photoprism.app/tensorflow/nsfw.zip
- photoprism-260601-a7d098548.tar.gz
- photoprism-import.service
- photoprism-import.timer
- photoprism-vision.service
- photoprism-vision.timer
- photoprism.service
- photoprism.sysusers
- photoprism.tmpfiles
- photoprism.user.service
aqua_rium commented on 2026-04-12 20:59 (UTC)
shtrophic commented on 2026-04-12 18:40 (UTC) (edited on 2026-04-12 18:42 (UTC) by shtrophic)
Hi fellow photo-selfhosting enjoyers, were back with split packages for tensorflow (and now, also onnxruntime), since photoprism won't work with updated libraries from extra.
To upgrade to 260305-3:
- install photoprism-onnxruntime-bin and photoprism-tensorflow-bin
- to pick the best version, you can refer to https://github.com/photoprism/photoprism/blob/df13ea9e024f815b8182d4e236f8e3e8ce6a722e/scripts/dist/install-tensorflow.sh#L67
- if installation fails because packages exist in the filesystem, try the pacman flag --override='*'.
- now update photoprism to 260305-3.
- if you had to use --override='*' for onnxruntime, you might need to reinstall photoprism-onnxruntime-bin for the files to actually be present.
Now everything should be in place for facedetection with onnx to run properly.
shtrophic commented on 2026-04-10 17:04 (UTC)
@aqua_rium care to share what you changed to make it work with pigo? I have had the issue with face detection as well but have not found the time to investigate...
aqua_rium commented on 2026-04-10 07:06 (UTC)
@shtrophic Face Recognition does not work out of the box with your build as PHOTOPRISM_FACE_ENGINE=auto defaults to onnx instead of pigo. Setting it to the latter works, but the needed resources for onnx (=auto) are missing. I found no clue how to add them, do you have an idea?
deepsea22 commented on 2025-12-04 09:22 (UTC)
@shtrophic Thank you! I didn't realize I needed to manually replace the config. It all works after removing the diff, specifically: 1a2
ModelsPath: /usr/share/photoprism 5a7 VisionYaml: /var/lib/photoprism/config/vision.yml
shtrophic commented on 2025-12-04 07:28 (UTC)
@deepsea22 please diff /etc/photoprism/defaults.yml{,.pacnew}
deepsea22 commented on 2025-12-03 14:21 (UTC)
After updating to latest, photoprism throws this error on startup: ERRO[2025-12-03T15:20:05+01:00] config: invalid models path, check configuration and permissions What am I doing wrong? I have not customized anything but the storage path, which worked fine previously.
krutoileshii commented on 2025-09-07 22:36 (UTC)
@AlynxZhou you know you can just overwrite the settings in the defaults.yml inside of the options.yml Since this package doesn't provide options.yml it should keep your settings between updates.
shtrophic commented on 2025-08-08 20:38 (UTC)
Thanks for noticing @Alynx/hou
AlynxZhou commented on 2025-08-08 13:22 (UTC)
@deepsea22 First install MariaDB and setup it (just follow Arch Wiki), and then create an user for PhotoPrism (I'll just use photoprism), set a password for it, then create a database for PhotoPrism (I'll call it photoprism too), and grant all permission for this user in this database.
And then you need to set following in options.yml to make PhotoPrism use MariaDB:
DatabaseDriver: mysql
DatabaseName: photoprism
DatabasePassword: your_database_user_password
DatabaseServer: /run/mysqld/mysqld.sock
DatabaseUser: photoprism
If your are using different machine for PhotoPrism and MariaDB, use IP:port for DatabaseServer instead of UNIX socket.
Pinned Comments
shtrophic commented on 2026-04-12 18:40 (UTC) (edited on 2026-04-12 18:42 (UTC) by shtrophic)
Hi fellow photo-selfhosting enjoyers, were back with split packages for tensorflow (and now, also onnxruntime), since photoprism won't work with updated libraries from extra.
To upgrade to 260305-3:
Now everything should be in place for facedetection with onnx to run properly.
evine commented on 2024-09-04 01:58 (UTC) (edited on 2024-09-24 04:22 (UTC) by evine)
Do not use nodejs 22.7.0/22.8.0 to make.