PyTorch-SSD

Single Shot Multi-Box Detector implementation in PyTorch.

This is the implementation used by SSDIR, a single-shot multi-object representation learning model.

Development

Requirements:

• Install poetry (https://python-poetry.org/docs/#installation)
• Use poetry to handle requirements
  • Execute poetry add <package_name> to add new library
  • Execute poetry install to create virtualenv and install packages

Training

To train the model use the train.py script. Activate the environment by running poetry shell and run python train.py --help to see all the available options.

See all the available datasets in datasets directory. To train on multiscale MNIST dataset generate the dataset using multiscalemnist tool.

A trained model weights file can be used for training SSDIR model.

Manual

Use make to run commands

• make help - show help
• make test - run tests
  • args="--lf" make test - run pytest tests with different arguments
• make shell - run poetry shell