Fisher Caffe

C++ implementation of learning Fisher vectors using VlFeat. In addition the fisher vector can be concatenated with another feature vector such as a Caffe layer etc.

Requirements

• CMake
• Boost
• Eigen
• OpenCV
• VlFeat

Training

• Extract Dense SIFTs from a set of labeled images
• PCA Project
• Compute GMM using EM
• Compute Fisher Vector for the images from the trained GMM
• Add additional features.
• Train one-against-many linear SVM for each class

Test

• Extract Dense SIFTs from the set of test images
• PCA Project
• Compute Fisher Vector for the images from the trained GMM
• Add additional features.
• Classify using linear SVM

This code is provided as is. If you notice any errors or bugs file a request or inbox me.