3D Reconstruction & Gaussian Splatting
Large reconstruction models, 3DGS scene representation, and the geometry/appearance trade‑offs that show up when you push them to real captures on real hardware.
Computer Vision · AR Glasses · 3D & 4D Reconstruction
Kunal A.
Kathare
I develop computer vision solutions for AR glasses — 3D and 4D reconstruction, Gaussian splatting, and depth completion models. The work runs from research to something that has to hold up on a device you wear.
Senior Computer Vision Research Engineer at Samsung R&D Institute India, Bangalore. Previously Vision & AI Lab, IISc.
- Publications
- 4
- Citations
- 7
- h‑index
- 1
- Venues
- WACV · ECCV
01 — About
I’m a Senior Computer Vision Research Engineer at Samsung R&D Institute India, Bangalore, where I develop vision solutions for AR glasses. My research sits in 3D and 4D reconstruction and depth completion — filtering a 4D Gaussian field so it holds up under zoom and off‑axis views, propagating sparse sensor depth into something dense enough to trust, and editing the triplane latents of a large reconstruction model with a sentence.
I first joined Samsung as a research intern in 2024 and returned full‑time in 2025. In between I worked at the Vision & AI Lab, CDS, IISc on 3D Gaussian splatting for reconstructing complex scenes. Before that I was at PES University, where I taught C programming labs and built an NLP‑driven indoor navigation robot that takes its instructions in Kannada.
I like problems where the failure mode is visible — where you can look at a render and immediately see what the method got wrong. Wearables make that unforgiving: the artefact is sitting on someone’s face.
02 — News
- Aug 26 Our paper Towards Alias‑Free 4D Gaussian Representations with Motion‑Aware Filtering was accepted to ECCV 2026.
- Jul 26 Hierarchical Convolution Spatial Propagation Networks for Depth Completion was accepted to SPCOM 2026.
- Jul 25 Joined Samsung R&D Institute India, Bangalore as a Computer Vision Research Engineer.
- Nov 24 NLP Enabled Autonomous Indoor Navigation Robot for Kannada was accepted to CSITSS 2024.
- Oct 24 Instructive3D was accepted to WACV 2025.
- Oct 24 Joined the Vision & AI Lab (VAL), CDS, IISc.
- Jan 24 Started as a research intern at Samsung R&D Institute India, Bangalore.
03 — Research interests
Dynamic & 4D Representations
Extending splatting into time — motion‑aware filtering, anti‑aliasing, and view synthesis that survives zoom and unusual viewpoints for AR/VR.
Generative & Text‑Guided Editing
Diffusion adapters in latent 3D space: changing pattern, colour and reflectance from a text instruction while geometry stays consistent.
Depth Completion
Hierarchical convolutional spatial propagation networks that turn the sparse, noisy depth a device actually measures into dense, usable predictions.
04 — Publications
2026
-
Towards Alias‑Free 4D Gaussian Representations with Motion‑Aware Filtering
ECCV 2026
2025
-
Instructive3D: Editing Large Reconstruction Models with Text Instructions
WACV 2025
2024
-
NLP Enabled Autonomous Indoor Navigation Robot for Kannada in Rural Health Centres
CSITSS 2024
05 — Experience
-
Jul 2025 – Present
Samsung R&D Institute India — Bangalore
Senior Computer Vision Research Engineer
Developing computer vision solutions for AR glasses: 3D and 4D reconstruction, Gaussian splatting, and depth completion models — the work behind the ECCV 2026 and SPCOM 2026 papers.
-
Oct 2024 – Jun 2025
Vision & AI Lab (VAL), CDS — Indian Institute of Science
Project Assistant
Research on 3D Gaussian splatting and its use in representing and reconstructing complex 3D scenes.
-
Jan 2024 – Sep 2024
Samsung R&D Institute India — Bangalore
Research Intern
Where the 3D work started — text‑guided editing of large reconstruction models in triplane latent space, which became Instructive3D at WACV 2025.
-
Jan 2024 – May 2024
PES University — Bangalore
Teaching Assistant, C Programming
Designed lab programming questions for the C programming course, and created and managed tests and assessments.
06 — Education
-
2020 – 2024
PES University — Bengaluru
Bachelor of Technology (B.Tech), Computer Science
CGPA 9.59 / 10
Final‑year work became the Kannada‑language autonomous indoor navigation robot published at CSITSS 2024. Teaching assistant for the C programming course in 2024.
-
2018 – 2020
SMPU — Dharwad
Class XII (Pre‑University), PCMS
94%
Scholarships & awards
- 2021 CNR Rao Scholarship — for outstanding academic performance in the 1st and 2nd semesters.
- 2022 MRD Scholarship — for academic excellence in the 3rd and 4th semesters.
- 2023 MRD Scholarship — for exceptional performance in the 6th semester.
07 — Get in touch