The 3D Scene Perception via Probabilistic Programming (3DP3) system developed by Massachusetts Institute of Technology researchers helps machines perceive the world in a more human-like manner.
The probabilistic programming framework enables the system to cross-check detected objects against input data, to determine whether camera-recorded images likely correlate with any candidate scene.
3DP3 can deduce whether mismatches are caused by noise or by errors in scene interpretation that require correction via further processing.
The system represents the scene as a graph, where each object is a node and the lines linking nodes indicate which objects contact one another; the result is more accurate estimations of object arrangement.
3DP3 almost always generated more accurate poses than other deep learning models, and performed significantly better when some objects were partially blocking others.
From MIT News
View Full Article
No entries found