why there are two inputs in RNN blocks?

zhjwustc / cvpr18_rnn_deblur_matcaffe

Dynamic Scene Deblurring Using Spatially Variant Recurrent Neural Networks[matcaffe]

MIT License

39 stars 9 forks source link

why there are two inputs in RNN blocks? #3

Open Akumio opened 6 years ago

Akumio commented 6 years ago

I looked at the code and found that rnn blocks contain GRUs in four orientations.One of the inputs is from the feature map after the slice, so how is the input after feature extraction entered? And what do these two inputs mean?

zhjwustc commented 5 years ago

One input is the RNN filtered features and the other is estimated RNN features.