Reproduce “Knowledge Graphs of Driving Scenes to Empower the Emerging Capabilities of Neurosymbolic AI” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ruwan Wickramarachchi, Cory Henson, Amit Sheth
The AI Institute at the University of South Carolina, in collaboration with the Bosch Center for Artificial Intelligence, developed DSceneKG, a suite of real-world driving scene knowledge graphs, to serve as a benchmark for neurosymbolic AI. These KGs, instantiated from autonomous driving datasets like NuScenes, comprise millions of triples and entities, supporting various complex AI tasks.
Key insights
- Formal ontologies can effectively structure complex, multimodal real-world data from diverse sources into usable knowledge graphs, bridging the gap between raw sensor data and symbolic knowledge.
- Large-scale, domain-specific knowledge graphs derived from industrial datasets provide a robust and realistic benchmark for evaluating and advancing neurosymbolic AI methods beyond theoretical or generalized contexts.
- Integrating explicit symbolic knowledge graphs with sub-symbolic learning enables advanced AI capabilities such as improved entity prediction, explainable reasoning, and semantic similarity computations in complex, dynamic environments.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.