I have been building the robot and playing around with mrl for a while now and everytime I attempted to setup face recognition, the programme would crash, as in the live video would pause and no mouse clicks would do anything. the speech still worked, but it would otherwise just hang. 
I initially thought it was the compuer I was using was too slow, however I have just updated to a new computer and same problem.

Is there something I am doing wrong or do I just need a massivly fast computer or what?

please any help would be good

kwatters

5 years 6 months ago

There is no reason that face recognition should crash anything. Do you have an error message? Check the MyRobotLab.log file for clues. Better yet, if you can reproduce the problem and send a no works! This shares the log files with us so we can help.

Are you using a script? If so can you share it? Also how many faces have you tried to train? Face recognition only works once you have trained at least 2 people.

Once a face is detected, the face recognition filter is going to make it's best guess on who it thinks it is...

it is when I try to train it that it stops. When I say crash, it basically freezes. so you cannot click on anything. I may be doing something wrong for all I know. I go in to open cv, ask it if it sees me, it turns on the camera and puts a square around my face, then when I click on the train it just stops. 

Am I supposed to put any parameters such as opencv face recognizer in or just leave it blank?

 

So, it's important to understand that when you click train, it is going to have to crunch through all the training images and compute a face recognition model.  This is an intensive task to compute.  It helps to limit the number of training images you have.

How many training images do you have per face?  Try limiting it to abouty 10 or 15..

Also, what type of computer are you using ?

I havn't got any training images in there. Do you need to upload them some other way? I thought that I was meant to click train to train it.

 

You should be interactively training the robot to create the training images.

The face recognizer has a train mode and a recognize mode.  When you're training, the face recognizer is creating and saving training/example images. 

When it's in recognize mode, it is using those images for classification in real-time.

From the Swing gui, you type in the name of the person you want to recognize.. then click the train button...  When a face is recognized (face+eyes+mouth) a training image will be saved into the mrl/training directory under a sub directory of the persons name.

 

You must have at least 2 people trained before the face recognizer will start attempting to recongize faces.  (it always makes a guess, is it person A or person B.. for example.)

 

Here's a video i did on it a while back.. it shouldn't be too far out of date...