DMI

Prerequisites

Environment

• Python 3.6.13
• PyTorch 1.10.1

  Installation

  Create a conda virtual environment and activate it

  conda create -n DMI python=3.6.12
  conda activate DMI
  pip install -r requirements.txt

  Data Preparation

  1. Dataset (SVD) is available here: SVD
  2. Create a folder ${DATASET_DIR} to store the features. The directory structure should look like this:

     ${DATASET_DIR}
     |--${QUERY}
       |-- ${FEATURES}
       |-- ${INDEX}
       |-- ${SHAPE}
     |--${LABELED}
       |-- ${FEATURES}
       |-- ${INDEX}
       |-- ${SHAPE}
     |--${UNLABELED}
       |-- ${FEATURES}
       |-- ${INDEX}
       |-- ${SHAPE}

     Start Training

     python main.py --cfg config/solver/DMI.yaml --log config/logging.yaml

     Start Test

     python main.py --cfg config/solver/baseline.yaml --log config/logging.yaml --resume ${CHECKPOINT}