MMM-Facial-Recognition-OCV3 error at tools.train.py

[SdeGeata]

@SdeGeata Anyone got an answer to this one? I’m still stuck on this.

[nischi]

@alfi153
Do you tried what are in the docu?

Test your trained model
Make sure your training.xml from running train.py is in this directory
Specify your user labels in the environment with

export FACE_USERS=name1,name2,name3

Run python tools.facerecognition.py to test your trained model.

[SdeGeata]

@nischi

Hello Nischi;

Only got as far as “python tools.train.py”. That is what generates the error and it doesn’t generate the XML file that it should, so I can’t verify my training.

Everything up until then is fine, and there are two named folders in the specified folder (training_data/) with pics of myself and my girlfriend. These photos were already processed by the previous function; “python tools.capture.py”. I don’t know why its bugging.

[nischi]

@SdeGeata
It seems he can’t resize the image for some reason. Have you checked all your pictures that nothing is corrupt and have you checked that your python has access to the files as well?

[SdeGeata]

@nischi

Yeah, all the photos load fine in preview app and they were all found, formatted and placed in the folder by the previous step. I don’t know why it is having a problem.

[SdeGeata]

@SdeGeata

I have gone ahead and started the process over again, just in case i missed something. I uninstalled everything and did a fresh install. It won’t even capture anymore, never-mind training with the images i already have.

I just want to get this thing going.

[nischi]

@SdeGeata
Do you working with a mac? Maybe you have some issues with the permissioning, have you tried (only for testing) to give all folders full access? i had some issue also if i used sudo for some stuff where i shoudn’t.

[nischi]

@SdeGeata
I am currently working on a new version of face recognition with another method and the new ocv 4 version. Current tests works much better than this old module. i only need to finish the module soon. i think this or next week i should have a first version which works.

[SdeGeata]

@nischi

Sorry, i just saw your replies…

I am working on a Mac via VNC viewer and Terminal, all the software is installed on the Rpi 3 b+ directly though. I followed the installation steps on the page thoroughly, so I don’t know what could have gone wrong.

I did since then install MotionEye for another aspect I want to add to the mirror and I notice that it adds “boxes” around detected movement - I don’t think it does this normally, but rather is related to the OCV install; only the recognition wouldn’t work anyhow due to the fact its not been “trained” who’s who.

Please let me know how your new version goes. I am desperate to get this facial recognition finally working on my mirror.

[SdeGeata]

@SdeGeata

Hey @Nischi! I got it going!!!

Two things fixed it. The first was a directory error that was preventing capture. Fixed with:

in capture.py:

from config import ToolsConfig -> from .config import ToolsConfig

the second was an outdated PIP that wasn’t getting updated without SUDO that was giving me the error on Training.

sudo pip install future --upgrade

Both are fixed now and XML is generated. Next on to configuration. Thanks to all for your help.

[nischi]

@SdeGeata
Cool that it works. I’m nearly done with my new module now. Will inform you if you can try out. And the good thing. The classes config from your old module will be compatible.

[SdeGeata]

@nischi

Cool. Lemme know.

[nischi]

@SdeGeata
Hi,
Now it’s finally done and tested on my own Magic Mirror with the PiCamera. Let me know if you have any issues and need some help.

[SdeGeata]

That’s great news! I’m gonna install it today and give it a whirl. I’ve spent too much time swearing about the other one not working like it should. I’ll let you know how it goes.

[SdeGeata]

Well, since my post yesterday, I’ve not got past building Open CV 4. Hangs at 20%, or 73%, or 100% and eventually freezes. it never finishes the compilation.

I completely wiped my SD card and did a fresh install of Rasbian - just to see if that would make a difference. So far, no go. I’m losing patience with this thing. It’s been well over a week that i’ve been fiddling around with various face recognition modules that kinda half work. i am anxious to try yours - i just can’t get past the Open CV bugs on install.

What should I do???

[nischi]

@SdeGeata
Sorry to hear that you have some issue to install OpenCV. Which Hardware do you have? On my Raspi 3+ it works fine, first run and all compiled.

Whats really important is that you extend your Swap, otherwise you will have some issues like you described.

Did you go through this documentation? https://www.pyimagesearch.com/2018/09/26/install-opencv-4-on-your-raspberry-pi/

sudo nano /etc/dphys-swapfile

# set size to absolute value, leaving empty (default) then uses computed value
#   you most likely don't want this, unless you have an special disk situation
# CONF_SWAPSIZE=100
CONF_SWAPSIZE=2048

sudo /etc/init.d/dphys-swapfile stop
sudo /etc/init.d/dphys-swapfile start

Note: Increasing swap size is a great way to burn out your Raspberry Pi microSD card. Flash-based storage have limited number of writes you can perform until the card is essentially unable to hold the 1’s and 0’s anymore. We’ll only be enabling large swap for a short period of time, so it’s not a big deal.

[SdeGeata]

@nischi

Hello again;

I finally got Open CV 4 installed successfully, I am re-installing my mirror modules now.

I did extend the swap to 2048, and that failed making -j4 and -j1. Finally I boosted the swap to 5000, it still failed on -j1, but it finally worked on -j2… go figure.

I was pretty tired of f**cking around trying to get that to install, it was pretty discouraging.

Thanks for your help and I will let you know how it goes once I get your new module installed.

[nischi]

@SdeGeata
Sorry to hear that you had such problems to install OpenCV. Hope it works now and let me know how it works.

[SdeGeata]

@nischi

Hello again;

So Open CV 4 installed, as well as all the other stuff specified on your module page, in the same versions.

I am trying to run the encode script:

python3 encode.py -i ../dataset/ -e encodings.pickle -d hog

but I keep getting the following error:

Traceback (most recent call last):
  File "encode.py", line 6, in <module>
    import face_recognition
ImportError: No module named 'face_recognition'

face_recognition is installed:

[nischi]

@SdeGeata hmmm thats really strange, in your terminal you can import it and it works and if you try to run my encode.py script it doesn’t.

stupid question, but i need to ask, sorry. do you run my encode.py also in the “dlib_test” virtual environment?