Marko Mihajlovic

PhD Student in Computer Vision @ETH Zurich. Prev research intern at @RealityLabs (Pittsburgh and Zurich).

[Joaquin Gajardo @JoaquinGajardoC]

Excited to announce that our paper GrowFields got accepted at ECCV 2026 🌱

We tackle the challenging task of 4D plant growth modeling with compositional neural velocity fields. Check https://t.co/eOgmdQYW9o for more details!

@marko_mih @serge_prokudin @SiyuTang3 @lukas_roth_1

[MrNeRF @janusch_patas]

Neural Texture Splatting: Expressive 3D Gaussian Splatting for View Synthesis, Geometry, and Dynamic Reconstruction

Contributions:
• We introduce Neural Texture Splatting (NTS), a novel representation that augments 3D Gaussian Splatting with neural per-primitive RGBA texture fields. This serves as a robust representation that enhances 3DGS-based methods across a wide range of tasks.

• We propose leveraging a global tri-plane network to predict localized texture fields. This enables expressive modeling of geometry and appearance while ensuring robustness across tasks.

• Extensive experiments demonstrate that our representation consistently improves existing state-of-the-art methods across diverse benchmarks, including dense and sparse-view novel view synthesis, surface reconstruction, and dynamic reconstruction.

[ELLIS @ELLISforEurope]

Meet @DoraKontog, Asst Prof at @DTUtweet 🇩🇰 and ELLIS Member.

Her work spans 3D perception & adaptive methods for autonomous agents.

“My goal is to build AI systems that continuously learn and adapt, & a curious team to explore these challenges together.”

#WomenInELLIS

[Marko Mihajlovic @marko_mih]

Our MVTracker Oral presentation is happening now in the Kalakaua Ballroom! #ICCV2025

[Frano Rajič @franorajic]

1/4 🚀 We’re excited to release MVTracker (ICCV 2025 Oral), the first data-driven multi-view 3D point tracker. MVTracker tracks arbitrary 3D points across multiple cameras, handling occlusions and varied camera setups without per-sequence optimization.

[Marko Mihajlovic @marko_mih]

📦 pip install VolumetricSMPL
🌐 https://t.co/wPGssT2CU6
📄 https://t.co/oIkxN6L5bm

📍 Join us at ICCV 2025 — Poster #465, Session 2
🗓️ Tuesday, Oct 21 | 3:15–5:15 p.m.
With collaborators @SiweiZhang13 @genligenli @kaifengzhao @LeaMue27 @SiyuTang3

[Marko Mihajlovic @marko_mih]

🚀 Excited to share VolumetricSMPL #ICCV2025 Highlight — a volumetric body model extending SMPL for efficient human reconstruction, motion, and interaction modeling.

Lightweight, differentiable, and supports SMPL, SMPL-H, SMPL-X out of the box.
🌐 https://t.co/wPGssT2CU6

[Matthias Niessner @MattNiessner]

📢📢Want to build 𝟑𝐃 𝐅𝐨𝐮𝐧𝐝𝐚𝐭𝐢𝐨𝐧 𝐌𝐨𝐝𝐞𝐥𝐬?📢📢

➡️We're looking for Diffusion/3D/ML/Infra engineers and scientists in Munich & London.

Get in touch and apply: https://t.co/atwcWtTyV5

#GenAI #foundationmodels #worldmodels #diffusion #transformers

[Zhenjun Zhao @zhenjun_zhao]

Multi-View 3D Point Tracking

Frano Rajič, @haofeixu, @marko_mih, @derek_siyuanli, Irem Demir, Emircan Gündoğdu, @leike_lk, @serge_prokudin, @mapo1, @SiyuTang3

tl;dr: per-view features+depth maps->fused 3D feature point cloud->kNN-based correlation->spatiotemporal transformer->overlapping windows->3D point trajectories

https://t.co/il8bbxZDEx

[Frano Rajič @franorajic]

1/4 🚀 We’re excited to release MVTracker (ICCV 2025 Oral), the first data-driven multi-view 3D point tracker. MVTracker tracks arbitrary 3D points across multiple cameras, handling occlusions and varied camera setups without per-sequence optimization.

[Marko Mihajlovic @marko_mih]

😍 So well deserved! Big congrats to @jianyuan_wang @davnov134 and the whole team, amazing work! 🚀

[SpAItial AI @SpAItial_AI]

Congrats to the #VGGT work for winning the best paper award at @CVPR!

Very proud to have @davnov134 our CTO @SpAItial_AI among the contributors -- congratulations!

[TheodoraKontogianni @DoraKontog]

🚨 Recruiting PhD students to join my team at @DTU_VisualComp & @AiCentreDK on 3D Vision!
🇩🇰 🚲🌊🏰 Copenhagen is an emerging hub for computer vision with a thriving community—and it’s an amazing city!
📢 Apply now: https://t.co/W6vtAuTDZY & please reach out with any questions!

[Marko Mihajlovic @marko_mih]

@janusch_patas At ECCV‘24 we introduced the SplatField method focusing on sparse reconstruction using neural fields with 3DGS. If you’re referring to ‘field splatting’ in that context. Here’s the link to your older post: https://t.co/v6IgMkPJ6q

[MrNeRF @janusch_patas]

SplatFields: Neural Gaussian Splats for Sparse 3D and 4D Reconstruction

"We observe that 3D Gaussian splats (3DGS) lack spatial autocorrelation of splat features, which leads to suboptimal performance in sparse reconstruction settings.

Our key idea is to introduce neural fields to regularize 3D gaussian splats for sparse 3D and 4D reconstruction.

We propose a novel optimization strategy that effectively regularizes splat features by modeling them as the outputs of a corresponding implicit neural field."

[MrNeRF @janusch_patas]

SplatFields: Neural Gaussian Splats for Sparse 3D and 4D Reconstruction

"We observe that 3D Gaussian splats (3DGS) lack spatial autocorrelation of splat features, which leads to suboptimal performance in sparse reconstruction settings.

Our key idea is to introduce neural fields to regularize 3D gaussian splats for sparse 3D and 4D reconstruction.

We propose a novel optimization strategy that effectively regularizes splat features by modeling them as the outputs of a corresponding implicit neural field."

[Marko Mihajlovic @marko_mih]

@DoraKontog @DTU_Compute @AiCentreDK Congrats Dora! 🎉🎉

[Marko Mihajlovic @marko_mih]

Happening now. Great work from colleagues at ETH 🎉

[Songyou Peng @songyoupeng]

"The straight line belongs to men, the curved one to God." — Antoni Gaudí

We will present 3D Edge Reconstruction from UDF at #CVPR2024 today at 10:30, Poster 162!

Amazing works done by Lei Li (great master thesis!), Zehao, @B1ueber2y, Remi & @mapo1!

https://t.co/0wKFwYiRsW

[Marko Mihajlovic @marko_mih]

This work is led by our outstanding master's student Xiyi Chen from @ETH_en.
w. @sfwang0928, @serge_prokudin, and @SiyuTang3.

[Marko Mihajlovic @marko_mih]

Excited to introduce Morphable Diffusion for consistent, controllable novel view synthesis of humans from a single image at #CVPR2024!

Join us at the poster session today at 10 am @CVPRConf.

Explore our project page & code: https://t.co/NL8tlEIkMU

[Marko Mihajlovic @marko_mih]

Zhiyin is presenting our 3D Gaussian Avatar at @CVPRConf . Drop by if you are interested in creating avatars from monocular videos @sfwang0928 @SiyuTang3 @AutoVisionGroup

Project page: https://t.co/SmchQPcWeC

[MrNeRF @janusch_patas]

Source code released: 3DGS-Avatar: Animatable Avatars via Deformable 3D Gaussian Splatting
https://t.co/h404acOeVt

[Marko Mihajlovic @marko_mih]

Join us at the poster session @CVPRConf if you are interested in learning more about compact motion representations.
@cnsdqzyz @SiyuTang3 @qianli_m @serge_prokudin

Project page: https://t.co/lNl2nlzAPb

[Marko Mihajlovic @marko_mih]

@songyoupeng @GoogleDeepMind Wow amazing view! Congratulations Songyou! 🎉🚀