echo -n myuser:mypasswd | base64 sudo kubectl apply -k cred/

Delete Images from the Private Docker Registry

Remove one image from your Docker Registry

Let's remove ubuntu 18.04 tag and the image from the Docker Registry. In order to do that we need to get the Docker-Content-Digest of the tag. Once you execute the command below, the sha value for ubuntu 18.04 tag will be returned to the console. You may copy that output.

$ curl -v --silent -H "Accept: application/vnd.docker.distribution.manifest.v2+json" -X GET http://localhost:5000/v2/ubuntu/manifests/18.04 2>&1 | grep Docker-Content-Digest | awk '{print ($3)}'

Now let's delete the image by using the Docker-Content-Digest ID which was returned.

$ curl -v -X DELETE http://localhost:5000/v2/ubuntu/manifests/<sha_value_you_received_from_the_previous_command>

Once you run this command you will receive a successful reply(status code 202).

Run the garbage collector

Now we need to run the Docker Garbage Collector to remove the image completely from the file system. From this step, you will be able to reduce the storage used.

$ docker exec localregistry bin/registry garbage-collect --delete-untagged /etc/docker/registry/config.yml

You will see that all the layers will be listed and only the related layers will get selected to be deleted.

PLEASE NOTE in registry version 2.7.1 :- At this moment you have successfully deleted your image. But if you again push the same image(Ubuntu 18.04) to your local Docker Registry, you will see a reply saying the image layers still exist. In order to avoid this you need to restart you Docker Registry.

$ docker restart localregistry

Now you can push the same image and you will see that its getting pushed.

Kubernetes with Kaniko

For build on kubernetes with Kaniko need change COPY in Dockerfile to assets/3dmm_assets and assets/pretrained_models

torch torchvision Python main / nightly main / nightly >=3.8, <=3.11 2.1 0.16 >=3.8, <=3.11 2.0 0.15 >=3.8, <=3.11 1.13 0.14 >=3.7.2, <=3.10

NextFace

Single Image

python3 optimizer.py --input input/2.jpg --output output

Batch for same face

python3 optimizer.py --sharedIdentity --input input --output output

mitsuba leak - couldn't get it to work. redner not support python 3.8 torch 1.13.1 not support 11.8, need python 3.9

HRN

tensorflow >2.11.0 need cudnn > 8.5.0 torch compiler not support cuda 11.8

python3.9 demo.py --input_type single_view --input_root ./assets/custom/input --output_root ./assets/custom/output
python3.9 demo.py --input_type multi_view --input_root ./assets/custom/input --output_root ./assets/custom/output

kubernetes build with kaniko - work, almost, need edit for folder structure and etc, not tested, deprecated.

python3.9 reconstruct_faces.py --input_type single_view --input_root ./assets/custom/input --output_root ./assets/custom/output

output['displacement_map'] (64, 64, 3) output['deformation_map'] (256, 256, 1)