search

davidsandberg / facenet

Face recognition using Tensorflow
MIT License
13.61k stars 4.8k forks source link

StackOverflow Question: Facenet/OpenFace realtime clustering #695

Open angelorodem opened 6 years ago

angelorodem commented 6 years ago

The question below was made by me on stackoverflow, can you help?

Recently i got interested in facial recognition and i read that that the OpenFace and Facenet fit a face in a 128-dimension sphere, and images of the same face tend to stay close to each other, in Openface docs they wrote about clustering and Facenet there is a python script related to clustering, but there very few explanations about how the whole clustering works.

Is it possible from a face detected from a live stream, fit it in the 128-d sphere, analyze to which cluster it belongs, being that this cluster is being populated in real-time as the photos are being fitted in the sphere, and is there some easy to understand example?

Thanks.

speculaas commented 6 years ago

Hi, Angelorodem, I think maybe https://github.com/davidsandberg/facenet/blob/master/contributed/cluster.py can help?

khadijakhaldi commented 5 years ago

did you figure out the answer ?

angelorodem commented 5 years ago

nope, i ended up with a solution that gave a incredible number of false positives, so i'm still waiting