Person Re-identification

This is a person re-identification (re-id) project that aims at learning generic deep features from multiple person re-id datasets.

Installation

We have integrated our self-brewed caffe into external/caffe, which provides batch-normalization and multi-gpu parallel training. Please clone this project with the command:

git clone --recursive https://github.com/Cysu/person_reid.git

Apart from the official installation prerequisites, we have several other dependencies: cudnn-v4, openmpi, and 1.55 <= boost < 1.60. You may install them manually or by a package manager (apt-get, pacman, yum, etc.).

Then configure the Makefile.config and compile the caffe.

cd external/caffe
cp Makefile.config.example Makefile.config
# Configure the libraries properly
make -j8 && make py

Download datasets

Download the following datasets.

1. CUHK03
2. CUHK01
3. PRID
4. VIPeR
5. 3DPeS
6. i-LIDS (I cannot find the link to the original dataset. This is my previous backup version.)
7. Shinpuhkan (need to send an email to the authors)

Link the root directory of these datasets to our project.

ln -sf /path/to/the/root/of/datasets external/raw

Prepare data

1. Create a directory for experiment data and results

   mkdir -p external/exp
   

   or link against another external directory

   ln -s /path/to/your/exp/directory external/exp
   

2. Convert raw datasets into a uniform data format

   scripts/format_rawdata.sh
   

3. Convert formatted datasets into LMDBs.

   scripts/make_dbs.sh
   

4. Merge all the datasets together for the joint single-task learning (JSTL).

   scripts/merge_dbs.sh
   

Experiments

Our experiments are organized into several groups:

1. Baseline: training individually on each dataset
2. Baseline: joint single task learning (JSTL)
3. Our method: JSTL with domain guided dropout (DGD)
4. Additional leave-one-out experiments

Baseline: training individually on each dataset

To train and test a model individually on a dataset, just run the script

scripts/exp_individually.sh prid

where the parameter is the dataset name, can be one of cuhk03, cuhk01, prid, viper, 3dpes, ilids.

Baseline: joint single task learning (JSTL)

First, pretrain a model using the mixed dataset with JSTL

scripts/exp_jstl.sh

After training, the script will use the pretrained model to do the evaluation directly on each individual dataset. The CMC accuracies printed out are corresponding to the JSTL entries in Table 3 of our paper.

Then, fine-tune the pretrained model on each dataset and evaluate the performance

scripts/exp_ft_jstl.sh

The CMC accuracies printed out are corresponding to the FT-JSTL entries in Table 3 of our paper.

Our method: JSTL with domain guided dropout (DGD)

Based on the pretrained JSTL model, we can compute the neuron impact scores (NIS) for each dataset, and resume the JSTL training with domain guided dropout.

scripts/exp_dgd.sh

The CMC accuracies printed out are corresponding to the JSTL+DGD entries in Table 3 of our paper.

Then, fine-tune the model on each dataset with domain guided dropout and evaluate the performance

scripts/exp_ft_dgd.sh

The CMC accuracies printed out are corresponding to the FT-(JSTL+DGD) entries in Table 3 of our paper.

Referenced Datasets

We summarize some commonly used person re-id datasets below.

Name	Reference
3DPeS	Baltieri, et al., 3DPes: 3D people dataset for surveillance and forensics
CUHK01	Li, et al., Human reidentification with transferred metric learning
CUHK02	Li, et al., Locally Aligned Feature Transforms across Views
CUHK03	Li, et al., Deepreid: Deep filter pairing neural network for person re-identification
i-LIDS	Zheng, et al., Associating groups of people
PRID	Hirzer, et al., Person re-identification by descriptive and discriminative classification
Shinpuhkan	Kawanishi, et al., Shinpuhkan2014: A Multi-Camera Pedestrian Dataset for Tracking People across Multiple Cameras
VIPeR	Gray, et al., Evaluating appearance models for recognition, reacquisition, and tracking