Export Reviews, Discussions, Author Feedback and Meta-Reviews

"NIPS Neural Information Processing Systems 8-11th December 2014, Montreal, Canada",,, "Paper ID:","664" "Title:","Encoding High Dimensional Local Features by Sparse Coding Based Fisher Vectors" Current Reviews First provide a summary of the paper, and then address the following criteria: Quality, clarity, originality and significance. This paper proposes a new Fisher vector coding method to encode high-dimensional local features. The proposed method assumes a generative model in which a local feature is drawn from a Gaussian distribution with a randomly generated mean vector. By approximating the likelihood, the resulting objective function becomes that of a sparse coding problem and the model parameters can be learnt by using the standard sparse coding solvers. A new Fisher coding vector can be derived by the differentiating the log likelihood of the generative model.