Learning Less is More – 6D Camera Localization via 3D Surface Regression: https://arxiv.org/pdf/1711.10228.pdf
MVSNet: Depth Inference for
Unstructured Multi-view Stereo: https://arxiv.org/pdf/1804.02505.pdf
Impressive. The source code will be published soon https://hci.iwr.uni-heidelberg.de/vislearn/research/scene-understanding/pose-estimation/#CVPR18
Seems a little shaky without the training input 3D model, it would be interesting to see how it performs with training over a set of nadir video shots from a drone.
MVSNet paper also looks really interesting, but no source code. I doubt they’ll publish anything, as some of the authors are associated with altizure.