yinhefeng Goto Github PK

followers: 64.0 following: 89.0 repos: 80.0 gists: 0.0

Name: Hefeng Yin

Type: User

Bio: Lecturer at Wuxi University, Wuxi, P.R.C

Hefeng Yin's Projects

real-world-masked-face-dataset

Real-World Masked Face Dataset，口罩人脸数据集

s3rc

Semi-Supervised Sparse Representation Based Classification for Face Recognition

saprocrc

Image Recognition with Sparsity Augmented Probabilistic Collaborative Representation based Classification

sbdcl

The Sparse Bayesian based joint Discriminative Dictionary and Classifier Learning algorithm

Supervised Sparse Coding With Decision Forest-IEEE Signal Processing Letters 2019 By jointly conducting sparse coding and classifier training, supervised sparse coding has shown its effectiveness in a variety of recognition tasks. However, the existing supervised sparse coding methods often consider linear classification, which limits their discrimination in handling highly nonlinear data. In this letter, we propose a new supervised sparse coding model by incorporating decision tree classifiers. Since decision trees can well deal with the non-linear properties of data, the introduction of decision trees to sparse coding can noticeably improve the discrimination of coding. Meanwhile, sparse coding is able to produce sparse de-correlated features that decision tree is in favor of. For further improvement, we close the loop of sparse coding and decision tree learning with an ensemble framework, which alternatively learns a dictionary for sparse coding and a decision tree for classification. The resulting series of decision trees as well as series of dictionaries are used to construct a decision forest for classification. The proposed method was applied to face recognition and scene classification, and the experimental results have demonstrated its power in comparison with recent supervised sparse coding methods.