Combining Unet with RNN

HI everyone,

In the course of “Sequence Models” by Andrew Ng, most the RNN is used for NLP purposes. Is there any simple reference where the RNN has been used for image classification? I have a medical image which has a kind of sequential behavior! The muscle and fat layers are repeated in 1 dimension. I have used Unet to classify the fat and muscle areas. Is it possible to use RNN for this 2D image (but similar to text the sequence behaviour is in 1D)? Do you think it will improve the accuracy? If I want to combine the Unet with the RNN, is there any architecture that you could suggest?

Thanks and regards,
Mohammad

Yes, it seems possible to use images as input to an RNN. I don’t have any suggestions for the architecture.