Open akturtle opened 6 years ago
R-FCN comes on top of object classification. So, it's practically redundant to apply it to the classification task. Simply use one of the pre-trained models in the MXNet model zoo for the classification task.
Have you successfully trained a classification network with deformable conv? @akturtle
R-FCN comes on top of object classification. So, it's practically redundant to apply it to the classification task. Simply use one of the pre-trained models in the MXNet model zoo for the classification task.