Datasets

• AirSim-MAP - Multi-agent perception dataset based on AirSim. Available here.
• PCB component detection datasets as part of our WACV 2019 paper is available here.

Code

We are strong proponents of reproducible research and open-source. Upcoming projects by the group will be released at our RIPL github.

General Frameworks we have released include:

• Continual learning framework as part of our NeurIPS Continual Learning Workshop paper.
• Framework for learning with weaker pair-wise information

Code for publications:

• ICLR 2021: Code for Unbiased Teacher for semi-supervised object detection
• NeurIPS 2020: Code for posterior re-calibration for imbalanced learning
• ECCV 2020: Code for FeatMatch for semi-supervised image classification.
• CVPR 2019: Code for action segmentation
• ICCV 2019: Code for our video domain adaptation method
• ICLR and CVPR 2019: Code for Self-monitoring navigation agent and Regretful navigation agent, achieving state of art on the Room 2 Room task
• ICLR 2019: Framework for learning with weaker pair-wise information
• ICLR 2019: Few-shot learning framework (collaboration with VT) can be found here.
• NIPS Continual Learning Workshop 2018: Continual learning framework as part of our NeurIPS Continual Learning Workshop paper.
• ICLR 2018: Neural network clustering and learning with pair-wise constraints in PyTorch (corresponding to our ICLR 2018 and 2019 papers). We also have older code from our ICLR 2016 workshop track paper in lua torch.
• DRAGAN: Notebook example for our DRAGAN paper.