Human reconstruction always limited to tight-fitting clothing👕? Not anymore! In our #ECCV2024 paper ReLoo, we propose a method to reconstruct the detailed geometry and appearance of humans dressed in loose garments💃from only monocular in-the-wild videos!🥳 1/7

Are you interested in hair reconstruction, human pose and shape estimation, digital human with clothing deformation, and hand-object interaction? Checkout our ECCV papers in the poster session at #ECCV2024! See 🧵.

Wow! Sapiens is nominated for ECCV best paper!! Presentation info: Oral 1C, Wed, 2nd Oct, 1:30 p.m. CEST, Silver Room. Poster 113. GitHub: github.com/facebookresear…

📢 Code is out! 📢 It runs end-to-end from an input video to the visualization below, no manual processing required! I will also be happy to chat about this work today at #ECCV2024 in the afternoon session: PS-6-#278

Had a lot of fun presenting our poster yesterday with Vanessa Sklyarova! 😃 Talked with many people interested in strand-based hair modelling, so hopefully we’ll see more works in this area in the future.

📢📢Introducing "Eliminating Oversaturation and Artifacts of High Guidance Scales in Diffusion Models" TL;DR: We show that with a few modifications to how the CFG update is applied, we can vastly mitigate the oversaturation and artifacts of high guidance scales. See the 🧵

Gaussian Haircut (#ECCV2024) featured on 2-minute papers: youtu.be/FyC5ZAMWDi4?si…

(1/5) Excited to announce our latest project: -> Joker: Conditional 3D Head Synthesis with Extreme Facial Expressions <- From one reference image we generate a 3D reconstruction with a novel extreme expression controlled with a 3DMM and text. Webpage: malteprinzler.github.io/projects/joker/

A purely synthetic face prior can generalize to casual in-the-wild captures and stylized faces 🤯 I'm happy to share our SIGGRAPH Asia ➡️ Hong Kong paper “Cafca: High-quality Novel View Synthesis of Expressive Faces from Casual Few-shot Captures.” syntec-research.github.io/Cafca/

Excited to present our recent work "LiteVAE: Lightweight and Efficient Variational Autoencoders for Latent Diffusion Models" next week at #NeurIPS2024. Please join us at our poster session next Thursday (December 12th) from 11:00 - 14:00 at East Exhibit Hall A-C #2604. 👇🧵

🚀 We present SynShot! A novel method leveraging synthetic prior to create fully drivable 3D head avatars using only a few shots (typically just 3 images).🎭✨ zielon.github.io/synshot arxiv.org/abs/2501.06903

📢Happy to announce that the code for Joker is now publicly available: github.com/malteprinzler/… Including pretrained weights and interactive demos 🤓 Feel free to play around!

Today is my last day at ETH Zurich!🥹 This has been a fantastic time. I found a lot of new collaborators and friends in and outside of Zurich. Thanks to all the members of the AIT lab for being incredible people to work with 😊 I'm excited about what comes next. Stay tuned!🤓

Meta presents: Pippo : High-Resolution Multi-View Humans from a Single Image Generates 1K resolution, multi-view, studio-quality images from a single photo in a one forward pass

📢📢 Last week, we announced Pippo - a DiT that generates 1K res. turnarounds from a single iPhone photo (even occluded ones)! Here’s the deep dive thread unpacking everything we learned! ⬇️

After some extra time and additional work, I am happy to announce that D3GA will be presented at 3DV2025 in Singapore. The code and dataset can be found below. 🚀🚀 zielon.github.io/d3ga/ github.com/facebookresear…

I’m happy to share that I’m starting a new position as a Research Scientist at Meta, Codec Avatars team! Working in this team has been my dream job since my early PhD days 😊 I'm super excited about what comes next!

🎉 I’m happy to share that 𝗯𝗼𝘁𝗵 𝗚𝗘𝗠 𝗮𝗻𝗱 𝗦𝘆𝗻𝗦𝗵𝗼𝘁 got accepted to 𝗖𝗩𝗣𝗥 𝟮𝟬𝟮𝟱! Check out the papers, code, and comparisons below: 🔹 GEM: github.com/Zielon/GEM 🔹 SynShot: github.com/Zielon/SynShot Thanks to all collaborators, and see you in Nashville! 👋

📢GeomHair: Reconstruction of Hair Strands from Colorless 3D Scans📢 We present a novel method to reconstruct hair strands from colorless 3D scans by extracting orientation cues directly from the mesh surface geometry by finding local characteristic lines and from shaded

Want Gaussian Avatar on mobile? Turns out the bottleneck is decoding of pose correctives. At #SIGGRAPH2025, we present a simple yet highly effective solution. We make *any* Gaussian avatars mobile-ready via linear distillation and corrective sharing. 👉forresti.github.io/squeezeme/