|
|
|
|
|
by muktabh
3218 days ago
|
|
To add a bit to Parth's answer it has been trained using metric learning type losses and normal softmax combined.
There is a next version of combined low/high level features combined algorithms we are slowly going to update into. |
|