search

nischi / MMM-Face-Reco-DNN

Face recognition with opencv and deep neural network
MIT License
91 stars 46 forks source link

Module not loaded #78

Closed evgenblum closed 3 years ago

evgenblum commented 3 years ago

Hello, I install everything but while loading the module I get the next errors- MM_errors

Can you help me please?

shivansh7299 commented 3 years ago

I got the same error please help...

nischi commented 3 years ago

Which Raspberry do you have and which camera do you have? Is the Camera working?

evgenblum commented 3 years ago

I use buster on PI 3 B+. Camera PI works fine and module MMM-MotionDetector based on this camera works.

evgenblum commented 3 years ago

When I wrote the previous post I thought maybe MMM-MotionDetector catch the camera and prevent loading your module. I remarked configuration MMM-MotionDetector and your module now loading. But I see a black screen and the magic mirror can't load. Screenshot_1

evgenblum commented 3 years ago

I got it, I forgot to set classes: 'always' for modules. Now I can see magic mirror screen. The last problem,- it can't detect my face. There is any clue to investigate why it can't detect my face?

nischi commented 3 years ago

did you create the pickle encoding eith a dataset?

evgenblum commented 3 years ago

yes. I have encodings.pickle Screenshot_2

evgenblum commented 3 years ago

May be I should rotate camera? What value should I set to rotateCamera for rotating 180 degrees?

nischi commented 3 years ago

phu not sure, try 😊

evgenblum commented 3 years ago

Yes, it's a rotated camera problem. Unfortunately, I can't find the right value for rotateCamera. I tried 0,1,2,3,180 - nothing helped. Only the physical rotation of the camera helps me to solve this problem. In my case, it's not the best solution.

nischi commented 3 years ago

not sure about the values. dis not this feature.

evgenblum commented 3 years ago

Maybe you know, there is a way to personalize global greetings from the compliments module? Or personalize your welcome message to show different texts for a specific person.

nischi commented 3 years ago

nope sorry

stefan-sherwood commented 3 years ago

You have to use the same rotation for the images you build the model with as when detecting. If your model doesn't include pictures in similar lighting it can also fail for that reason too.