I want to predict a pixel given n pixels to it's left and n pixels to its right, using bi-directional RNN. Would i need to pass the 2n pixels to the RNN and look at the very last output of the RNN to get my prediction? Or pass 2n + 1 pixels, including the pixel i want to predict as 0, placed in the middle of those n and n pixels, and look at the middle unit output of the RNN for the prediction of that pixel?
I would like to suggest you some links: e.g., http://www.easy-tensorflow.com/tf-tutorials/recurrent-neural-networks/bidirectional-rnn-for-classification
https://www.oreilly.com/ideas/using-tensorflow-to-generate-images-with-pixelrnns
Read this paper:
http://users.cecs.anu.edu.au/~Tom.Gedeon/conf/ABCs2018/paper/ABCs2018_paper_92.pdf
Read this paper:
http://users.cecs.anu.edu.au/~Tom.Gedeon/conf/ABCs2018/paper/ABCs2018_paper_92.pdf
- Badges
- Science topic
- Similar topics
- Computer Science
- Images