search

FlyEgle / MAE-pytorch

Masked Autoencoders Are Scalable Vision Learners
245 stars 36 forks source link

How to apply "mae" to a detection pipeline? #10

Closed zobeirraisi closed 2 years ago

zobeirraisi commented 2 years ago

Thanks for sharing the code; I have a question or help me with how I can use mae as a backbone for the object detection framework? If you have any guidance, I'd appreciate it if you could please help me with that?

FlyEgle commented 2 years ago

Mae is only the pretrain, it also is build on the VIT model which have Encoder and Decoder, remove the decoder, U can use the vit encoder as the backbone and add the detection head to make detection.