davidsandberg / facenet

Face recognition using Tensorflow
MIT License
13.83k stars 4.81k forks source link

How To Train MTCNN #214

Closed herleeyandi closed 7 years ago

herleeyandi commented 7 years ago

Hello I am new in Tensorflow and deep learning. Now I am in face alignment task. I want to train MTCNN using my own dataset. So anybody here can give me some hint to do that?, for example:

  1. Format of the dataset, like naming the folder and image.
  2. Train the network to produce the model
  3. Test one face image and output the result.
  4. Is there ane pretrained model?, if exist, how I used it?

Maybe some of you will think this basic ones in tensorflow, but really I don't get any idea to train this MTCNN, so any other hint and suggestion will be appreciate. Thank you, sorry my English is bad.

yekeping commented 7 years ago

@herleeyandi it may be not right,but I just change the folder of the author,the accuracy is also my problem.why do I follow the code of the author,but the accuracy is low on my dataset

herleeyandi commented 7 years ago

I see, now it make sense. OK I am agree 100% with you. So how can I train with your MTCNN Code?, hint, suggestion, and step?

sun9700 commented 7 years ago

Will @davidsandberg release the code of training facenet (with Triplet Loss) and MTCNN on CASIA-WebFace and MS-Celeb-1M ?

davidsandberg commented 7 years ago

No such plans right now.