search

espressif / esp-dl

Espressif deep-learning library for AIoT applications
MIT License
548 stars 118 forks source link

Please Help!!!! - frmn.c #41

Closed AD121390 closed 3 years ago

AD121390 commented 3 years ago

I need a copy of frmn.c. I am using esp-face on my esp32 for a school project, and my professor wants my group to do the facial recognition part. I know that face_id has the feature extractions that I need for facial recognition. I just do not know what those 512 elements are measuring.

Any help is appreciated. Thank you.

TiramisuJ commented 3 years ago

Hi, This 512-dimensional vector is the high-dimensional feature of the face. we did not constrain the specific face information represented by each dimension. But in a high-dimensional space, the distance between two vectors of the same person is very small, while the distance between two vectors of different people will be very large.

AD121390 commented 3 years ago

I am having a problem where the face_id vector is the same no matter what face is detected by the esp32 camera. How can I fix this???

TiramisuJ commented 3 years ago

First of all, make sure that the image obtained by the camera is correct, and then use the API we provide to extract the features of the face. You can refer to this https://github.com/espressif/esp-who/tree/master/examples/single_chip/face_recognition_with_command_line to write your code