PVL Library Error : "FaceRecognizer: internal database is full."

[Akash_K_]

Hi,

I am using the OpenCV Face Recognition Sample Application which is bundled as part of Intel OpenVINO windows version – “computer_vision_sdk_2018.3.343” SDK.

The problem we are facing is as follows:

1. Although the code suggests that there should be no restrictions on the number of trained faces in the DB which is currently limited to 320.

Defined in facerecognizer.cpp

//--------------------------------------------------------------------------------

// define: editable

//--------------------------------------------------------------------------------

//Maximum number of db member for faces

//Unlimited, but recommend less than 1000 for better performance

#define MAX_FACES_IN_DATABASE       320

2. The issue we face is, as we try a register more than 32 faces we see a DB full error thrown by the library.

[pvl_face_recognition_register_facedata:547] PVL_EXCEPTION(pvl_err_database_full) from IntelFaceRecognizer_RegisterFace_Feature(fr, facedata->data, facedata->person_id, facedata->face_id)

OpenCV(3.4.2-openvino_2018_R2.0.0) Error: Internal error (Assertion: err != pvl_err_database_full failed. "FaceRecognizer: internal database is full.") in cv::pvl::FaceRecognizerImpl::registerFace, file C:\jenkins\workspace\OpenCV\OpenVINO\build\openvino_extra\pvl\src\face_recognizer.cpp, line 375

OpenCV: terminate handler is called! The last OpenCV error is:

OpenCV(3.4.2-openvino_2018_R2.0.0) Error: Internal error (Assertion: err != pvl_err_database_full failed. "FaceRecognizer: internal database is full.") in cv::pvl::FaceRecognizerImpl::registerFace, file C:\jenkins\workspace\OpenCV\OpenVINO\build\openvino_extra\pvl\src\face_recognizer.cpp, line 375

We are using the Face Recognizer sample project under the OpenCV samples directory. Ideally we want to have a capability to register and save around 300 faces as stated by the application.  

Could Anybody help me on this issue.

Regards,

Akash

[Gilnam_C_Intel]

hi,

The maximum number of registerable 'person' is 32. And each person can have 10 faces which means you can register 10 faces with same person ID. So the maximum number of registerable 'face' is 320.

Use below APIs to get exact number

getMaxRegistrablePersons()  --> return 32
getMaxRegistrableFacesPerPerson() --> return 10
getMaxRegistrableFaces() --> return 320 ( 32 x 10)

[Akash_K_]

Hi Choi,

Thanks for replying back. 

So the maximum number of faces we register is 32, that I understood. There is another parameter "MaxRegistrableFacesPerPerson", How I can use this if i want to train 10 faces for a Person.

Regards,
Akash