Skip to content

A Systematic Evaluation and Benchmark for Person Re-Identification: Features, Metrics, and Datasets

License

Notifications You must be signed in to change notification settings

HoYinTam/person-reid-benchmark

 
 

Repository files navigation

Person Re-identification Benchmark

This repository hosts the codebase for the following work: Karanam, S., Gou, M., Wu, Z., Rates-Borras, A., Camps, O., & Radke, R. J. (2016). A Systematic Evaluation and Benchmark for Person Re-Identification: Features, Metrics, and Datasets. IEEE Transactions on Pattern Analysis and Machine Intelligence, accepted February 2018.

Tested on Windows Server 2012 with MATLAB 2016b

Quick Start

  • Clone this repository
  • Run a quick example in run_experiment_benchmark.m
  • Read the results for VIPeR dataset with WHOS feature and XQDA

Run other experiments

  • Download supported dataset, unzip it and put it under the folder ./Data
  • Download corresponding partition file and put it under the folder ./TrainTestSplits
  • Run corresponding prepare_DATANAME.m inside the folder ./Data (if avaliable)
  • Change the parameters in run_experiment_benchmark.m

Check List for supported/tested feature

  • HistLBP
  • WHOS
  • gBiCov
  • LDFV
  • ColorTexture\ELF
  • LOMO (Windows)
  • GOG (Windows)

Check List for supported/tested metric learning

  • FDA
  • LFDA
  • kLFDA-linear/chi2/chi2-rbf/exp
  • XQDA
  • MFA
  • kMFA-linear/chi2/chi2-rbf/exp
  • NFST
  • KISSME
  • PCCA-linear/chi2/chi2-rbf/exp
  • rPCCA-linear/chi2/chi2-rbf/exp
  • kPCCA-linear/chi2/chi2-rbf/exp
  • PRDC
  • SVMML
  • kCCA

Check List for supported/tested multi-shot ranking method

  • rnp
  • srid
  • ahisd

Check List for supported/tested dataset

Reference

About

A Systematic Evaluation and Benchmark for Person Re-Identification: Features, Metrics, and Datasets

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • HTML 60.3%
  • C 14.3%
  • MATLAB 12.7%
  • C++ 7.0%
  • Fortran 2.0%
  • Python 1.0%
  • Other 2.7%