MART v0.2.1

We fixed a performance bottle neck in #19 that should make the training of object detection models 3x faster.

You can install v0.2.1 by running $ pip install https://github.com/IntelLabs/MART/archive/refs/tags/v0.2.1.zip.

MART v0.2.0

We change the way MART searches for configuration YAML files:

1. If a configs folder is not in the current working directory, MART would search in mart/configs from the package installation folder.
2. If a configs folder is in the current working directory, MART would search configuration in the order of ./configs and mart/configs.

In this way, users can run MART from any folder after installation. A configs folder is not required any more.

You can install v0.2.0 by running $ pip install https://github.com/IntelLabs/MART/archive/refs/tags/v0.2.0.zip.

v0.1

Initial release of the Modular Adversarial Robustness Toolkit (MART). MART makes it easy to compose novel attacks to evaluate adversarial robustness of deep learning models.