These are chat archives for beniz/deepdetect

25th
Jun 2016
Qiang Guo
@ih4cku
Jun 25 2016 08:29
Why does this project named "detect"?
It's not for detection.
Emmanuel Benazera
@beniz
Jun 25 2016 08:59
How does classification differs from detection exactly ?
Qiang Guo
@ih4cku
Jun 25 2016 09:00
Detection may involved with regression which is different from classification.
Emmanuel Benazera
@beniz
Jun 25 2016 09:01
and ? you can do regression with DD if you like
however, the true answer is that detection is a form of binary classification
and DD is in use for many detection tasks, though it supports classification and regression which are tasks with larger scope
Emmanuel Benazera
@beniz
Jun 25 2016 09:08
@ih4cku and if you are looking for object detection, there are ways to do it with fully convolutional nets such as resnets, which are supported.
Qiang Guo
@ih4cku
Jun 25 2016 09:09
Thanks for the explaination, Emmanuel. What's the difference of using DD with directly use Caffe?
Does DD support recurrent network?
Emmanuel Benazera
@beniz
Jun 25 2016 09:20
Dd is for production. See the related issue for RNN / LSTM, it's just been merged into Caffe.
Also we use our own Caffe, with some improvements, see the readme and links therein
Qiang Guo
@ih4cku
Jun 25 2016 09:21
Thanks! I'll see the docs. :)