Moving Towards Realistic Symmetry-Based Completion

DURATION:

5 min

FORMAT:

Presentation

Computer vision applications rely on 3D geometry recognition and point cloud completion. But relying on deep neural networks (DNN) requires large complex data sets that can be hard to collect.

Leverage your Symmentry-Based Completion

Our speaker:

Taras Rumezhak, R&D Engineer, SoftServe

Taras Rumezhak presents a novel framework that doesn’t rely on training data and outperforms SOTA methods with superior results on real-world datasets.