Openface error while training using docker

[vijay krishna]

HI All,

Based on the below link am trying to train the dataset (LFW downloaded from web)
https://cmusatyalab.github.io/openface/demo-3-classifier/

While running the below command encountered the error message.
th main.lua -outDir /features -data   /training

Error Message

Updating classList and imageClass appropriately   
/root/torch/install/bin/luajit: /root/openface/batch-represent/dataset.lua:219: Class has zero samples: Kultida_Woods
stack traceback:
    [C]: in function 'error'
    /root/openface/batch-represent/dataset.lua:219: in function '__init'
    /root/torch/install/share/lua/5.1/torch/init.lua:91: in function </root/torch/install/share/lua/5.1/torch/init.lua:87>
    [C]: in function 'dataLoader'
    /root/openface/batch-represent/batch-represent.lua:19: in function 'batchRepresent'
    main.lua:42: in main chunk
    [C]: in function 'dofile'
    /root/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:145: in main chunk
    [C]: at 0x00406670

Pls suggest the way forward.

Thanks
vijay

[Antonio Jesus Rodriguez]

Hi, this is because after running the alignment either the folder was empty or wasn't able to align detect any face in the pictures in the folder. In your case, Kultida_Woods. There is an script in the util folder, called prune-dataset which you can use to remove empty directories or directories with less than N files.

You can use as an alternative this too for removing empty folders:

$ find . -type d -empty -delete

[vijay krishna]

Thanks Antonio.... my issue is resolved.

one more question pls, I am using the standard docker (

docker bamos/openface)container for the below implementation
After system restart i lost all the changes in the container. 
Please suggest how i copy the standard docker container content
to my personal docker container and move the same to production.

Pls suggest the way forward.

[Antonio Jesus Rodriguez]

Hi,

As answered on private:

 To solve that you need to start docker with a shared volume so it uses your files from the local machine instead of the local image.

I do it like this:

sudo docker run -p 9000:9000 -p 8001:8001 -v /path-do-my-folder/docker_data:/root/openface/docker_data -t -i bamos/openface /bin/bash

so docker uses the images that I have in my local docker_data folder.

I'm following the next tutorial: https://medium.com/@ageitgey/machine-learning-is-fun-part-4-modern-face-recognition-with-deep-learning-c3cffc121d78#.p8b4iqqgy

El sábado, 11 de febrero de 2017, 12:44:59 (UTC+1), vijay krishna escribió:

[stephen.n...@gmail.com]

On Saturday, 11 February 2017 19:01:26 UTC+3, Antonio Jesus Rodriguez wrote:

Hi, this is because after running the alignment either the folder was empty or wasn't able to align detect any face in the pictures in the folder. In your case, Kultida_Woods. There is an script in the util folder, called prune-dataset which you can use to remove empty directories or directories with less than N files.

You can use as an alternative this too for removing empty folders:

$ find . -type d -empty -delete

This definitely solved it. Thanks