Deep learning for 3D understanding of satellite images

In this episode, Miko hosts 3 guests to talk about using deep learning for 3D understanding of satellite images: Dawa Derksen, Roger Marí, and Yujiao Shi. The episode covers a summary of each guest’s contributions on the topic as well as a panel discussion.
Timeline:
(00:00:05) Episode Introduction
(00:01:52) Guest Introduction
(00:06:43) Dawa Derksen: Origins of Shadow-NeRF
(00:28:55) Roger Marí: EO-NeRF
(00:47:48) Yujiao Shi: Connecting Satellite Image with StreetView
(01:13:58) Discussion
---
Guests:
Dawa Derksen - Dawa received the master’s degree in aerospace engineering from the Institut Supérieur de l’Aéronautique et de l’Espace (ISAE-Supaéro), Toulouse, France, in 2016, and the Ph.D. degree from the the Centre d’Etudes Spatiales de la Biosphère (CESBIO) Laboratory, Toulouse, France, in 2019. His Ph.D. topic was the operational production of image processing algorithms applied to the large scale classification of Earth Observation images for land cover mapping. Dr. Derksen pursued a post-doctoral research fellowship at the European Space Agency from 2020-2022, and is currently working at the Centre National d’Etudes Spatiales (French Space Agency) where he is involved in the field of 3D Implicit Representation Learning applied to Remote Sensing.
Shadow-NeRF: github.com/esa...
Roger Marí - Roger is a post-doc researcher from Barcelona specialized in 3D vision tasks (reconstruction, calibration, co-registration, change detection). He is currently working at the Centre Borelli, ENS Paris-Saclay, in France, where his research topic is the application of neural rendering methods to satellite image collections. He is the author of Sat-NeRF and EO-NeRF, some of the first models in the literature to provide quantitatively convincing results in terms of surface reconstruction.
Homepage: rogermm14.gith...
Sat-NeRF: centreborelli....
EO-NeRF: rogermm14.gith...
Yujiao Shi - Yujiao is a research fellow at the Australian National University. She obtained her PhD degree at the same institute. Her research interests include satellite image-based localization, cross-view synthesis, 3D vision-related tasks, and self-supervised learning. She has published eight first-authored papers on satellite image-related tasks in top-tier conferences, including CVPR, ICCV, TPAMI, NeurIPS, AAAI, ACCV, etc. She was also invited as a tutorial speaker on satellite image-based localization at CVPR 2023 and is a workshop co-organizer on UAV-based localization at ACM MM 2023.
Homepage: shiyujiao.gith...
"Boosting 3-DoF Ground-to-Satellite Camera Localization Accuracy via Geometry-Guided Cross-View Transformer": arxiv.org/abs/...
"CVLNet: Cross-View Semantic Correspondence Learning for Video-based Camera Localization": arxiv.org/abs/...
"Accurate 3-DoF Camera Geo-Localization via Ground-to-Satellite Image Matching": arxiv.org/abs/...
"Geometry-Guided Street-View Panorama Synthesis from Satellite Imagery": arxiv.org/abs/...
---
Host & Production: Mikolaj Czerkawski mikonvergence....

[@읽기쉬운마음]

love you

[@psy_duck8221]

Thank you for the wonderful podcast and top insights. The music was a bit distracting though.

[@jimj2683]

Can't they just train an AI model on all the Google Street View images and their locations (with input from a database containing local satellite images, maps, local photos etc). Then the model could guess the interpolated views between the Street View images and you will be able to move around freely.