University of Maryland computer science researchers are addressing a challenge in machine learning: teaching AI systems to recognize objects as they change shape, like sliced fruits. Their project, Chop & Learn, creates a dataset that enables these systems to identify produce in various states, including being peeled, sliced, or chopped. The team recorded themselves preparing 20 types of fruits and vegetables in different styles and from various angles to develop a comprehensive dataset.
This research, presented at the 2023 International Conference on Computer Vision in Paris, aims to improve AI's ability to understand objects undergoing transformation. This advancement could have applications in 3D reconstruction, video generation, and enhancing safety features in autonomous vehicles, and may even contribute to developing a robotic chef in the future. The project highlights the ongoing efforts to bridge gaps in AI's visual recognition capabilities.
Source