Applying Knowledge Distillation for Instance Segmentation using Mask RCNN (Tensorflow and Keras)

Hello sir,

I have a project related to instance segmentation.

I have done Mask RCNN for the dental dataset ( panoramic x-ray images to segment the teeth). I have two classes for the labels ( permanent and primary teeth).

I want to apply the knowledge distillation to my problem. The teacher model has been done but I need to do the student model and see how to evaluate the outputs of the two models.

Anyone can help and have knowledge or experience about KD ( teacher student model)?

Thanks