your paper and implementation is not the same?

luoyetx / deep-landmark

Predict facial landmarks with Deep CNNs powered by Caffe.

BSD 3-Clause "New" or "Revised" License

282 stars 156 forks source link

your paper and implementation is not the same? #34

Closed KangolHsu closed 7 years ago

KangolHsu commented 7 years ago

you said absolute value rectification(tanh) and locally sharing weights is efficient in facial point detection in your paper,but i did not see them in this implementation.could you give me some clues?

luoyetx commented 7 years ago

I'm not the author of the paper