search

terryum / awesome-deep-learning-papers

The most cited deep learning papers
25.49k stars 4.48k forks source link

Added entry to classic papers list: "Hierarchical Neural Networks for Image Interpretation #17

Closed sven-behnke closed 7 years ago

sven-behnke commented 7 years ago

In the thesis "Hierarchical Neural Networks for Image Interpretation", which was published as LNCS 2766 by Springer in 2003, I propose a hierarchical, recurrent, convolutional architecture for learning image interpretation. Many features of deep learning architectures popular today were present there, including output at original image resolution, supervised training through rectifying nonlinearities, and iterative refinement of interpretations. The main contribution is the use of horizontal and vertical feedback loops (via recurrent connections) that allow for the flexible use of contextual information for resolving local ambiguities.

terryum commented 7 years ago

Hi, Sven. I'm sorry, but, class papers are chosen among the papers which are cited more than 1000 times. I'm sorry.

Best regards, Terry