training

[3073]

How that is possible when you say single image per person is sufficient to train the algorithm??

[aangfanboy]

It is not when you are training, i used CASIA-WebFace dataset for traing. But this model is different than regular classification models. When you give a face image to model, model gives 512 numbers to you as output, which means 512 features. And in testing base, i give a image of let's say Elon Musk, model gives 512 features and i save them to my database. And when i give second image of Elon Musk to model, the features will be really close since it is the same person. This is what single shot means.

[3073]

Thank you

On Sat, Jan 18, 2020 at 12:31 PM Burak notifications@github.com wrote:

It is not when you are training, i used CASIA-WebFace dataset for traing. But this model is different than regular classification models. When you give a face image to model, model gives 512 numbers to you as output, which means 512 features. And in testing base, i give a image of let's say Elon Musk, model gives 512 features and i save them to my database. And when i give second image of Elon Musk to model, the features will be really close since it is the same person. This is what single shot means.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/aangfanboy/TripletLossFace/issues/2?email_source=notifications&email_token=AIC6A5LLW2FP7BTKPM2GNK3Q6LD6DA5CNFSM4KISAVS2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEJJUJZY#issuecomment-575882471, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIC6A5JWJGTWTOFYMRLPIPTQ6LD6DANCNFSM4KISAVSQ .

-- tttttggggggttttttg

[aangfanboy]

You're welcome