Talking Papers Podcast

KeypointNeRF - Marko Mihajlovic

October 19, 2022 Itzik Ben-Shabat Season 1 Episode 15
Talking Papers Podcast
KeypointNeRF - Marko Mihajlovic
Show Notes

 In this episode of the Talking Papers Podcast, I hosted Marko Mihajlovic . We had a great chat about his paper "KeypointNeRF: Generalizing Image-based Volumetric Avatars using Relative Spatial Encoding of Keypoints”, published in ECCV 2022.

In this paper, they create a generalizable NeRF for virtual avatars. To get a high-fidelity reconstruction of humans (from sparse observations), they leverage an off-the-shelf keypoint detector in order to condition the NeRF. 

Marko is a 2nd year PhD student at ETH, supervised by Siyu Tang. His research focuses on photorealistic reconstruction of static and dynamic scenes and also modeling of parametric human bodies. This work was done mainly during his internship at Meta Reality Labs. Marko and I met at CVPR 2022. 

Marko Mihajlovic, Aayush Bansal, Michael Zollhoefer, Siyu Tang, Shunsuke Saito

 Neural implicit fields have recently emerged as a useful representation for 3D shapes. These fields are Coordinate-based networks have emerged as a powerful tool for 3D representation and scene reconstruction. These networks are trained to map continuous input coordinates to the value of a signal at each point. Still, current architectures are black boxes: their spectral characteristics cannot be easily Image-based volumetric humans using pixel-aligned features promise generalization to unseen poses and identities. Prior work leverages global spatial encodings and multi-view geometric consistency to reduce spatial ambiguity. However, global encodings often suffer from overfitting to the distribution of the training data, and it is difficult to learn multi-view consistent reconstruction from sparse views. In this work, we investigate common issues with existing spatial encodings and propose a simple yet highly effective approach to modeling high-fidelity volumetric humans from sparse views. One of the key ideas is to encode relative spatial 3D information via sparse 3D keypoints. This approach is robust to the sparsity of viewpoints and cross-dataset domain gap. Our approach outperforms state-of-the-art methods for head reconstruction. On human body reconstruction for unseen subjects, we also achieve performance comparable to prior work that uses a parametric human body model and temporal feature aggregation. Our experiments show that a majority of errors in prior work stem from an inappropriate choice of spatial encoding and thus we suggest a new direction for high-fidelity image-based human modeling.


💻Project website
📚 Paper

To stay up to date with Marko's latest research, follow him on:
👨🏻‍🎓Personal Page
👨🏻‍🎓Google Scholar


If you would like to be a guest, sponsor or just share your thoughts, feel free to reach out via email: talking.papers.podcast@gmail

🎧Subscribe on your favourite podcast app:

📧Subscribe to our mailing list:

🐦Follow us on Twitter:

🎥YouTube Channel: