Skip to content

Latest commit

 

History

History
61 lines (53 loc) · 1.71 KB

README.md

File metadata and controls

61 lines (53 loc) · 1.71 KB

STMN

This repository contains a Pytorch implementation for our STMN (ICCV 2021). Our code is released only for scientific or personal use. Please contact us for commercial use.

Requirements

  • Python 3.6.8
  • Pytorch 1.1.0
  • Cuda 10.0
  • Cudnn 7.5
  • Pillow
  • progressbar2
  • tqdm
  • pandas

Getting Started

Datasets

  • Create your own database files, MARS for example, as follows:
cd database
python create_MARS_database.py \
    --data_dir 'path/to/MARS/' \
    --info_dir /path/to/MARS dataset/MARS-evaluation/info/ \
    --output_dir ./MARS_database/

Train

  • You can train our model using the below commands. Note that, in advance, you have to change variables 'TRAIN_TXT', 'TRAIN_INFO', 'TEST_TXT', 'TEST_INFO', and 'QUERY_INFO' in train.sh according to which dataset you want to use for the triaining.
cd smem_tmem
sh train.sh

Test

  • You can test a pre-trained model using the below commands. Similarly, you have to change variables 'TRAIN_TXT', 'TRAIN_INFO', 'TEST_TXT', 'TEST_INFO', and 'QUERY_INFO' in test.sh according to which dataset you want to use for evaluation.
  • Specify the path to pre-trained model parameters using 'LOAD_CKPT'
cd smem_tmem
sh test.sh

Acknowledgements

Our code is inspired by STE-NVAN