This repository contains a group of functions in C++ for training deep networks with Caffe:
-
dataset_reiden_multishot: generation of pairs from Re-Identification datasets with the division of the samples in train, validation and test sets for the multi-shot re-identification problem described in [1]. Additionally, the code allows several strategies for obtaining multiple images from one of the cameras.
-
Dataset used: PRID 2011
These functions follow a similar structure of the single-shot code: data_factory_from_reid, avalaible at https://github.com/magomezs/dataset_factory.
Please cite dataset_reiden_multishot in your publications if it helps your research:
José Héctor Penadés-Migallón, Re-identificación de personas a partir de múltiples capturas mediante aprendizaje automático,
Trabajo Fin de Grado, Universidad Carlos III de Madrid, July 2020.
This is an example of how to use data_factory_from_reid_multi with PRID2011 [2]:
string prid= "PRID_DATASET_DIRECTORY";
read_initial_parameters(argc, argv, &seed, &number_of_b_samples, &stride, &random_b_samples);
get_multiple_samples(prid, 7,4, number_of_b_samples,stride, random_b_samples);
for(int i=1; i<number_of_b_samples+1; i++);
train_val_test_division_multiple(prid, 100, 100, 100, 10, 100, 649, 100, &tag, i);
create_pair_data_multiple(prid, 100000, 10000, 1, 1, 1);
for(int i=2; i<number_of_b_samples+1; i++)
create_pair_data_multiple_remaining(prid, i);
tag=true;
for(int i=1; i<number_of_b_samples+1; i++)
create_test_data_multiple(prid, i, &tag);
NOTE: be careful with PRID samples whose identification number is higher than 200, because different people in cam a and b are labbelled with the same number, from id 200. Alternative solution: remove samples with ID higher than 200 in cam_a set, they are not neccesarry in the training and test described in [3].
[1] José Héctor Penadés-Migallón, Re-identificación de personas a partir de múltiples capturas mediante aprendizaje automático, Trabajo Fin de Grado, Universidad Carlos III de Madrid, July 2020.
[2] M. Hirzer, C. Beleznai, P.M. Roth, and H. Bischof, “Person Re-ID (PRID) Dataset”. Institute of Computer Graphics and Vision, 2011. Avalaible at: https://www.tugraz.at/institute/icg/research/team-bischof/lrs/downloads/prid11/
[3] M. Hirzer, C. Beleznai, P.M. Roth, and H. Bischof (2011). Person re-identification by descriptive and discriminative classification. In Scandinavian conference on Image analysis, pages 91–102. Springer.