Where is flatten? #2988
-
|
Where is a |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
|
If you're talking about that functionality in the dnn part, there isn't because it's not needed. Usually, you use flatten in TF to input the tensor into a Fully Connected (or Dense, as TF calls it) layer. In dlib you can input the tensor without flattening, and it will still work. See any Flattening is a human construct, the data remains the same, we just need to connect all neurons in FC layers, no matter where they are in space. |
Beta Was this translation helpful? Give feedback.
-
|
great, that's what I understood from the code |
Beta Was this translation helpful? Give feedback.
If you're talking about that functionality in the dnn part, there isn't because it's not needed. Usually, you use flatten in TF to input the tensor into a Fully Connected (or Dense, as TF calls it) layer. In dlib you can input the tensor without flattening, and it will still work. See any
dnn_example.Flattening is a human construct, the data remains the same, we just need to connect all neurons in FC layers, no matter where they are in space.