The 2022 AAAI Symposium on Machine Learning and Knowledge Engineering for Hybrid Intelligence (AAAI-MAKE) took place on March 21-23. The symposium is aimed at bringing together researchers and practitioners that are exploring the integration of Machine Learning methods with knowledge-based systems, to improve the explainability, performance and data efficiency of existing solutions.
This year, the symposium was run as a hybrid event, with the option to attend in person at Stanford University or to join from remote, via Zoom. Agnese presented her work titled “Robots with Commonsense: Improving Object Recognition through Size and Spatial Awareness” remotely. The work presented in this paper is part of her PhD research, supervised by Enrico Motta and Enrico Daga. In this paper, a novel method is proposed to extract knowledge of the typical sizes and spatial relations of objects from large-scale, general-purpose Knowledge Bases. Experiments on images collected by the HanS robot in KMi, which depict 60 object classes commonly found in our Lab, show that the introduced awareness of size and spatial knowledge can significantly augment object recognition methods that are solely based on Deep Learning.
In addition to the robotic applications that Agnese has presented, this year’s program covered a variety of use-cases, including, for example, applications geared towards the healthcare domain, Augmented Reality platforms designed for manufacturing domains, and human-in-the-loop solutions aimed at explaining Law cases. The program and proceedings of the symposium are available at https://www.aaai-make.info/program/