Question about the project

[so-park]

Hi @povilasb

I'm working on researching PCA for face recognition and I came across your codes.
It is very neat and I would love to use it, however, I'm not able to train the program for some reason.
It stops after "enter the name: " part. I also wasn't able to press keys to take another picture, it seemed like the train part terminates after a certain period of time. When I try to demo it, it says "not enough values to unpack". I made sure that I install all the dependencies, do you have any idea why this would be happening? Or any suggestions as to how I can fix this problem and use the program?

Thank you so much!!
I attached a screenshot as well. Thank you!

[povilasb]

I haven't tested it for ages! Let me see :)
Anyway, keep in mind that this is a very toy project. Personally eigenfaces didn't work very well for me in practice. It's a simple algorithm that compares face similarity pixel by pixel. So even a change of lighting will have a huge impact. See the previous discussions: #1 (comment)

[povilasb]

mm, I should have created requirements.txt with specific dependencies. From what I see, numpy has changed it's API and you're probably using latest numpy.
Here's a list of dependency versions I've been using:

cycler==0.10.0
decorator==4.0.11
matplotlib==2.0.0
networkx==1.11
numpy==1.12.1
olefile==0.44
Pillow==4.1.0
pyparsing==2.2.0
python-dateutil==2.6.0
pytz==2017.2
PyWavelets==0.5.2
scikit-image==0.13.0
scipy==0.19.0
six==1.10.0

See if installing exactly those versions help. If so, I'll commit requirements.txt file. Feel free to do so yourself and create a PR :)

[so-park]

Hello! Thank you for the reply! I honestly didn’t expect to get answer from you because the repository hasn’t been actively updated for a while. But really really thank you for writing me back. I really appreciate the help! Unfortunately, I have different versions of dependencies. Would it affect how the project is working? It still activates webcam but automatically takes ten pictures and doesn’t store any pictures. I only need a quick demo to show how face-recognition works PCA and it would be really nice if I can get it to work! Sincerely, Sion Park

…

On Feb 1, 2019, at 2:48 AM, Povilas Balciunas ***@***.***> wrote: mm, I should have created requirements.txt with specific dependencies. From what I see, numpy has changed it's API and you're probably using latest numpy. Here's a list of dependency versions I've been using: cycler==0.10.0 decorator==4.0.11 matplotlib==2.0.0 networkx==1.11 numpy==1.12.1 olefile==0.44 Pillow==4.1.0 pyparsing==2.2.0 python-dateutil==2.6.0 pytz==2017.2 PyWavelets==0.5.2 scikit-image==0.13.0 scipy==0.19.0 six==1.10.0 See if installing exactly those versions help. If so, I'll commit requirements.txt file. Feel free to do so yourself and create a PR :) — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#7 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AfwHj2lTsVkhdnXPQU28YsIUu1kK7WmMks5vJBt7gaJpZM4adm7->.

[povilasb]

mmm, my webcam currently is broke, I can't really test it. If by any chance you're at https://fosdem.org/ this weekend, let's meet ;) Otherwise I'll get some webcam and debug this later :) But sure, I'm pretty positive the dependencies can be updated relatively easy.

[so-park]

I see. It looks like a pretty cool conference!! But I won’t be there as I’m studying in the States haha. Yes please let me know!!! I have about a month, I will be looking for other resources in the meantime. Hope you get a lot from the conference and have a great time. Thank you again for being willing to help!

…

On Feb 2, 2019, at 4:16 AM, Povilas Balciunas ***@***.***> wrote: mmm, my webcam currently is broke, I can't really test it. If by any chance you're at https://fosdem.org/ <https://fosdem.org/> this weekend, let's meet ;) Otherwise I'll get some webcam and debug this later :) But sure, I'm pretty positive the dependencies can be updated relatively easy. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#7 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AfwHj6EKWqwU6H6vxVVq_7hvCuh_88RJks5vJYGUgaJpZM4adm7->.

[povilasb]

Hey, sorry for the delay, busy with the rest of my life...
Anyway, just tested the implementation. It works, although it's fragile. So

1. make sure you train your face before running demo
2. when training terminates too fast, that means you don't position your camera towards your face. See

   face-recognition/facerecognition/main.py

   Line 14 in 24a4c08

   for face in webcam.capture().faces():

   - train command immediately takes an image and tries to detect face in it. If no face is detected within an image, it will simply retry. It will attempt to take 10 images but all of the retries can fail.
3. you can play around with code interactively, e.g.

img = webcam.capture()
img.show()

for face in webcam.capture().faces():
    face.show()

[so-park]

Thank you! I was able to take 10 pictures and run demo, however, it looks like only the first picture is saved to the database. I see ten lines of “loop” but it only prints “taken” once in the beginning. Let me know if you have suggestions! Thank you for taking time to go over it.

…

On Feb 26, 2019, at 11:35 AM, Povilas Balciunas ***@***.***> wrote: Hey, sorry for the delay, busy with the rest of my life... Anyway, just tested the implementation. It works, although it's fragile. So make sure you train your face before running demo when training terminates too fast, that means you don't position your camera towards your face. See https://github.com/povilasb/face-recognition/blob/24a4c083bc69ec11fefc16aaa417339f32d4fb30/facerecognition/main.py#L14 <https://github.com/povilasb/face-recognition/blob/24a4c083bc69ec11fefc16aaa417339f32d4fb30/facerecognition/main.py#L14> - train command immediately takes an image and tries to detect face in it. If no face is detected within an image, it will simply retry. It will attempt to take 10 images but all of the retries can fail. you can play around with code interactively, e.g. img = webcam.capture() img.show() for face in webcam.capture().faces(): face.show() — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#7 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AfwHjz5Phuoqh7pcYvnItfaYltwLxIEcks5vRYxpgaJpZM4adm7->.