Reconstructing environments using SLAM скачать в хорошем качестве

Reconstructing environments using SLAM

Jingwen Wang gives a talk on Neural Implicit Representations in Simultaneous Localisation and Mapping (SLAM). Neural implicit representation has become more and more popular since the explosion of Neural Radiance Field (NeRF). It has shown its advantages and amazing results in various applications such as novel view synthesis, multiview stereo, offline surface reconstruction from multiview images, etc. However, very few previous works discuss the application of neural implicit representations in the context of SLAM, i.e. reconstructing the surrounding environments in an online sequential manner from a moving camera. Jingwen is a PhD student at UCL Foundational AI CDT under the supervision of Prof. Lourdes Agapito working on SLAM and 3D vision. He is interested in the intersection of Deep Learning and traditional SLAM, including object level SLAM, semantic scene understanding, neural scene representation.