Code for CVPR2020 Paper "Diverse Slow-Drop Block Multi-Branch Network for Person Re-Identification"

This code is developed based on pytorch framework.

Prerequisites

• Pytorch(1.1.0)
• yacs (0.1.6)
• torchvision(0.3.0)
• tb-nightly(2.0.0)
• Cython(0.29.12)
• pytorch-ignite(0.1.2)

SetUp

git clone https://github.com/freeSubmission/SDB.git

Data Preparation(Duke)

• dataset
  • dukemtmc-reid
    • DukeMTMC-reID
      • bounding_box_train
      • bounding_box_test
      • query

Train

cd SDB/scripts
vim train.sh
bash train.sh

Results(no rerank)

	rank 1	mAP
MARKET 1501	95.9	88.7
DUKE	90.2	79.5
CUHK-labeled	80.8	78.7
CUHK-detected	77.9	75.2

Datasets

• Market1501 https://www.cv-foundation.org/openaccess/content_iccv_2015/papers/Zheng_Scalable_Person_Re-Identification_ICCV_2015_paper.pdf`
• CUHK03 <https://www.cvfoundation.org/openaccess/content_cvpr_2014/papers/Li_DeepReID_Deep_Filter_2014_CVPR_paper.pdf>
• DukeMTMC-reID <https://arxiv.org/abs/1701.07717>