Skip to content

Latest commit

 

History

History
66 lines (46 loc) · 2.59 KB

README.md

File metadata and controls

66 lines (46 loc) · 2.59 KB

HPatches MATLAB

MATLAB implementation of the HPatches benchmark with code to reproduce the results in [1].

For the list of available commands run hb help.

Requirements

For running the HBenchmark MATLAB interface you will need MATLAB >R2016a.

If you do not have a MATLAB license (>R2016a), you can use the command line interface from a binary package. However to run the binary package, you need to install the MATLAB Compile Runtime:

More details how to install the MCR can be found here. Please note that around 2GB of free space is required.

Reproduce Article Results

To reproduce article results, download the used descriptor csv files and run:

run_article; res_article;```

This will generate figures in `results/article` (pgf/tikz format).
Scores for each descriptor are stored in `scores/scores_all`.


<a id=matlab-interface></a>
### MATLAB Interface

Main access point is the `hb` command from MATLAB:

```matlab
hb COMMAND ...

For more details of available commands, run hb help.

Command line interface

To run the HBench command line interface on Linux with the MCR located in the default path, run:

./bin/run_hb.sh /usr/local/MATLAB/MATLAB_Runtime/v901 COMMAND ...

If you have MATLAB installed e.g. in /usr/local/MATLAB/R2016a, you can use that instead of the MCR:

./bin/run_hb.sh /usr/local/MATLAB/R2016a COMMAND ...

You can see the list of all available commands here.

1: Vassileios Balntas, Karel Lenc, Andrea Vedaldi and Krystian Mikolajczyk: HPatches: A benchmark and evaluation of handcrafted and learned local descriptors. In CVPR 2017.