About
Klemen Kotar is a Researcher at the Allen Institute for Artificial Intelligence (AI2), focusing on Embodied AI, Computer Vision, and Reinforcement Learning. His work includes developing frameworks like AllenAct and analyzing self-supervised learning pipelines.
- Quick Links: Allen Institute for AI, Semantic Scholar, Twitter, Github
His research aims to enhance embodied agents' ability to reason about physical environments and improve lifelong learning. Recent work spans speech representation, Spelke object segmentation, and physical dynamics modeling with state-of-the-art results on the Physion benchmark.
Notable publications include AllenAct (2020), a PyTorch-based framework for embodied AI, and Contrasting Contrastive Learning (2021), analyzing over 700 experiments in self-supervised vision encoders. His work also covers interactive LEGO assembly simulation and adaptive object detection in embodied environments.
Collaborations include researchers like Roozbeh Mottaghi and Luca Weihs, with contributions to projects such as iTHOR, RoboTHOR, and Habitat. A Github repository showcases his technical implementations.
