docker-compose occurs error that manifest unknown

[jjaegii]

I used the wizard to create a docker-compose.yml file for Jetson and run docker-compose up command. but "ERROR: manifest for gokalpgoren/face_recognition_pytorch:latest not found: manifest unknown: manifest unknown" has been occured. I think it happened because there was no image of the arm version. So, can you upload the arm version of onvf, face_recognition_pytorch on your docker hub?

[mehmetgoren]

Hi,

Yes, you're right - there isn't an ARM version of the PyTorch face recognition image. I'll work on creating one next week.

To access the ARM version of the Onvif image, please use the following link: https://hub.docker.com/layers/gokalpgoren/onvf/v1.0.0-arm64v8/images/sha256-5b6e8d502326965ffadeb4d9ce83710946e63ea779c4a9ba33d79c779a6560b0?context=repo

[jjaegii]

Thank you!

[mehmetgoren]

I believe creating an Arm image for PyTorch face recognition is impractical due to its high hardware requirements, particularly a Nvidia GPU. Instead, I recommend utilizing the Deepstack face recognition service, which is better suited for Arm-powered devices. This approach will provide efficient and effective face recognition capabilities without the need for excessive hardware power.

[jjaegii]

thx! it worked but i can't use login and register functions

[Petrox]

I believe creating an Arm image for PyTorch face recognition is impractical due to its high hardware requirements, particularly a Nvidia GPU. Instead, I recommend utilizing the Deepstack face recognition service, which is better suited for Arm-powered devices. This approach will provide efficient and effective face recognition capabilities without the need for excessive hardware power.

There are Orin and Xavier AGXs with 64Gb ram and others with 90+ cores, so theoretically ARM could be relevant.

Nonetheless it's not necessary for the maintainer to build all variants, maybe just provide a script to do so.

How to build all arm64 images: (depends on two-three pullrequests to be merged)

#!/bin/bash

for module in feniks.web_app ffmpeg_service mngr senseai_service smcp snapshot_service ; do
 cd $module
 #echo "Building $module"
 echo -n "cd $module ; "
 command=`grep arm64 Dockerfile.run |grep build`
 echo -n "$command ;"
 echo "cd .."
 cd ..
done