POP-3D: Open-Vocabulary 3D Occupancy Prediction from Images - EXAKI training

EXAKI training number 1.

In this session, we present a joint work produced in collaboration between CIIRC at CTU in Prague and valeo.ai within the EXA4MIND project.

This work focuses on leveraging multimodal learning, particularly language-vision integration, to enable open-vocabulary 3D scene understanding. This is based on our work called “POP-3D: Open-Vocabulary 3D Occupancy Prediction from Images”, which was presented at NeurIPS.