Adapting NeRF for Low-Latency Reconstruction
Machine Learning Research @ Brown University (NSF-REU)
During this NSF-REU, I worked under the supervision of James Tompkin to develop a novel ENerf-backboned pipeline to optimize instantaneous 2D→3D data conversion for VR teleoperation, achieving a ~110° field-of-view. I conducted camera calibration for Boston Dynamics’ Spot robots and performed baseline tests. I also improved robot arm motion control using inverse kinematics through MR headsets. I presented this work at Leadership Alliance National Conference 2023 and Tanner Conference 2023.
Github repo: https://github.com/mlu108/enerf-scene-reconstruction
