This is a PyTorch implementation of the Exemplar Streaming (ExStream) algorithm from our ICRA-2019 paper. An arXiv pre-print and the published version of our paper are available.

• Python 3.5, PyTorch 1.0.0, NumPy, SciPy, scikit-learn, NVIDIA GPU
• Dataset:
  • CUB-200 -- ResNet-50 embeddings (included in cub200_resnet) version version

To generate results for the CUB-200 experiment with capacities [2,4,8,16] for the iid and class iid data orderings:

• python run_exstream_experiment.py

To generate plots for each of the experimental results:

• python plot_results.py

Our original paper used class-specific buffers for storing exemplars. In this implementation, you can maintain class-specific buffers by setting the buffer_type parameter to class; we also give the option to maintain a single buffer that will fill to full capacity and then begin replacement/merging by setting the buffer_type parameter to single

• [1] 'The Caltech-UCSD Birds-200-2011 Dataset.' by C. Wah, S. Branson, P. Welinder, P. Perona, and S. Belongie, http://www.vision.caltech.edu/visipedia/CUB-200-2011.html

If using this code, please cite our paper.

@inproceedings{hayes2019memory,
  title={Memory Efficient Experience Replay for Streaming Learning},
  author={Hayes, Tyler L and Cahill, Nathan D and Kanan, Christopher},
  booktitle={International Conference on Robotics and Automation (ICRA)},
  year={2019},
  organization={IEEE}
}