(1/n)
📢📢 𝐍𝐞𝐑𝐒𝐞𝐦𝐛𝐥𝐞 𝐯𝟐 𝐃𝐚𝐭𝐚𝐬𝐞𝐭 𝐑𝐞𝐥𝐞𝐚𝐬𝐞 📢📢

Head captures of 7.1MP from 16 cameras at 73fps:
* More recordings (425 people)
* Better color calibration
* Convenient download scripts

github.com/tobias-kirsc...

Excited to announce ScanNet++ v2!🎉

@awhiteguitar.bsky.social & Yueh-Cheng Liu have been working tirelessly to bring:
🔹1006 high-fidelity 3D scans
🔹+ DSLR & iPhone captures
🔹+ rich semantics
Elevating 3D scene understanding to the next level!🚀

w/ @niessner.bsky.social
kaldir.vc.in.tum.de/scannetpp

(3/3)
We further introduce a new multiview dataset of native English speakers with overall recording time of ∼3.5 hours.

shivangi-aneja.github.io/projects/gau...
youtu.be/2VqYoFlYcwQ

Great work by Shivangi Aneja, Artem Sevastopolsky, @tkirschstein.bsky.social , Justus Thies, @adai.bsky.social

Reminder: I will present GGHead at #SIGGRAPHAsia2024 on Friday in the Gaussian Humans session, starting at 10:45am in hall B5 (2). See you tomorrow!

(4/4)
As 3D Gaussians are sensitive to adversarial training, we introduce tailored regularization strategies and a UV total variation loss to facilitate high geometric fidelity of the generated heads.

Collaboration with Simon Giebenhain, Jiapeng Tang, Markos Georgopoulos and @niessner.bsky.social

(3/4)
We learn to generate photo-realistic 3D heads only from training on 2D images. In contrast to existing works such as EG3D, we leverage the efficient rasterization of 3D Gaussians, thus mitigating the need for any 2D super-resolution networks.
-> enforcing view consistency.

(2/4)
We generate photo-realistic 3D heads and render them with Gaussian Splatting at 1k resolution in real-time.

Also see our project page: tobias-kirschstein.github.io/gghead/
and video: youtu.be/M5vq3DoZ7RI

(1/4)
I will present GGHead next week at SIGGRAPH Asia in Tokyo! Swing by in the Gaussian Humans session if you are interested.

Also, we have released the training code now 🎉
Check it out: github.com/tobias-kirsc...