Vehicle Classification

This code is for vehicle classification using DNN features.

Instruction

Run demo

Before running the code:

• Download this project to your computer.

• Install VLFeat for generate fisher vector.

• LIBSVM is already installed in reference folder. You may follow the instructions if you want to compile on your computer.

• Run run.m in script folder, then wait for a while, you can get the result.

Generate DNN feature by yourself

If you want to generate DNN feature by yourself, follow the following procedure:

• Download the vehicle dataset.

• Modify the code in set_dir.m so that [dirs.vehicle] points to the vehicle dataset. You may also set a new [dirs.feature] to where you want to store the generated feature.

• Install Caffe for feature extraction on Alexnet. Also you need to install Matcaffe.

• Download caffemodel for Alexnet and save it to your corresponding caffe model directory. Also copy the deploy_single.prototxt to the caffe model directory.

• Modify the corresponding code in set_dir.m, regarding the directory of your caffe root directory.

• Run run.m in script folder, then you can get the generated feature and accuracy result.