Machine Learning and Computer Vision expert. Former Assistant Professor at the Department of Engineering, University of Cambridge
Cambridge, UK
Joined October 2011
- Tweets196
- Following2.3K
- Followers509
- Likes122
Ignas Budvytis retweeted
Locally controlling SMPL body shape has uses in 3D shape estimation and generation. Check out the following repo for body-part-based modification of SMPL shape. There is a lot of room for research and improvements!
Code: github.com/akashsengupta1997…
Ignas Budvytis retweeted
Easily one of the best talks I’ve heard at #CVPR2024!
Ignas Budvytis retweeted
NPLMV-PS: Neural Point-Light Multi-View Photometric Stereo
Fotios Logothetis, @IgnasBud, @robertocipolla
tl;dr: pixel intensity->3D shape; model point light attenuation, explicitly raytrace cast shadows and optimise a fully neural material renderer
arxiv.org/pdf/2405.12057
Ignas Budvytis retweeted
Task Agnostic Architecture for Algorithm Induction via Implicit Composition. arxiv.org/abs/2404.02450
Ignas Budvytis retweeted
I'm excited to introduce our work on efficient visual localization - PRAM: Place Recognition Anywhere Model for Efficient Visual Localization with @IgnasBud @robertocipolla
1. Self-defined 3D landmarks as in humans’ perception system - not limited to classic semantic labels
Ignas Budvytis retweeted
PRAM: Place Recognition Anywhere Model for Efficient Visual Localization
Fei Xue, @IgnasBud, @robertocipolla
tl;dr: map-centric landmark definition; recognition->sparse self-defined landmarks->landmark-wise matching
new paradigm for visual localization?
arxiv.org/pdf/2404.07785.pdf
Ignas Budvytis retweeted
Excited to be presenting “DiffHuman: Probabilistic Photorealistic 3D Reconstruction of Humans” at #CVPR2024!
Project page: akashsengupta1997.github.io/…
Huge thanks to my awesome @GoogleAI internship hosts: @thiemoall, @nikoskolot, @enric_corona, Andrei Zanfir, @CSminchisescu!
1/n
Ignas Budvytis retweeted
Excited to be presenting my latest work "Sparse Multi-Object Render-and-Compare" at #BMVC2023. CAD model based 3D scene representations have exciting applications for AR and robotics.
- with @IgnasBud and @robertocipolla
youtu.be/UFWL92rbyIU
Ignas Budvytis retweeted
very interesting work. similar idea to our CVPR2023 paper IMP: Iterative Matching and Pose Estimation with Adaptive Pooling w/ @IgnasBud @robertocipolla
arxiv.org/pdf/2304.14837.pdf
Source code including the training part is available at github.com/feixue94/imp-rele…
LightGlue: Local Feature Matching at Light Speed
@PhilippCSE, @pesarlin, @mapo1
tl;dr: exit mechanism+point prune between each layer in SuperGlue->adaptive stopping mechanism->fast inference of SuperGlue
github.com/cvg/LightGlue
arxiv.org/pdf/2306.13643.pdf
Ignas Budvytis retweeted
Can't wait to be at #CVPR in Vancouver next week! Drop in to the PM poster session on Tuesday to check out HuManiFlow - our probabilistic approach to 3D human pose and shape estimation.
Video: youtube.com/watch?v=6xDiJNzP…
More details: akashsengupta1997.github.io/
![[CVPR '23] HuManiFlow: Ancestor-Conditioned Flows on SO(3) for Human Pose and Shape Prediction](https://pbs.twimg.com/card_img/2114050994099109888/aogDK4gV?format=jpg&name=800x419)
Ignas Budvytis retweeted
Introducing our work 'IMP: Iterative Matching and Pose Estimation with Adaptive Pooling': efficient matching and pose estimation with transformers.
#CVPR2023
w/ @IgnasBud @robertocipolla
paper: arxiv.org/abs/2304.14837
code: github.com/feixue94/imp-rele…
Ignas Budvytis retweeted
@CVPR A new approach of using sematics for long-term localization
SFD2: Semantic-guided Feature Detection and Description @feixu94 @IgnasBud @robertocipolla
github.com/feixue94/sfd2
implicit semantics embedding into keypoints; no gt for training; no explicit labels at test time
Ignas Budvytis retweeted
Really looking forward to presenting our paper "HuManiFlow: Ancestor-Conditioned Normalising Flows on SO(3) Manifolds for Human Pose and Shape Distribution Estimation" at CVPR 2023!
w/ @IgnasBud and @robertocipolla
Paper + code: github.com/akashsengupta1997…
1/3
Ignas Budvytis retweeted
SFD2: Semantic-guided Feature Detection and Description
Fei Xue, @IgnasBud, @robertocipolla
tl;dr: more principled approach to use segmentation for image matching, than class-name filtering.
arxiv.org/abs/2304.14845.pdf
#CVPR2023
Ignas Budvytis retweeted
Introducing IronDepth, a framework that uses surface normal and its uncertainty to iteratively refine the predicted depth map (to appear in #bmvc2022). Visit baegwangbin.github.io/IronDe… for more detail.
Joint work with @IgnasBud and @robertocipolla.
Ignas Budvytis retweeted
Happy to share our latest work (with @BaeGwangbin, @IgnasBud and @robertocipolla) "SPARC: Sparse Render-and-Compare for CAD model alignment from a single RGB image" which will be presented at #BMVC2022 in November. #Sparse #3dshape
Ignas Budvytis retweeted
I am happy to share that our paper "SPARC: Sparse Render-and-Compare for CAD model alignment in a single RGB image" will be presented at #BMVC2022.
arxiv: arxiv.org/pdf/2210.01044.pdf
authors: Florian Langer, @BaeGwangbin, @IgnasBud, @robertocipolla
1/2
Ignas Budvytis retweeted
I am happy to share that our paper "Multi-View Depth Estimation by Fusing Single-View Depth Probability with Multi-View Geometry" will be presented at #CVPR2022 as an oral (top 4%)! Here is the link to the presentation.
youtu.be/LM113ibJVmQ
1/n
![[CVPR 2022 Oral] MaGNet](https://pbs.twimg.com/card_img/2115250786674388992/dGrmcnrW?format=jpg&name=800x419)
Ignas Budvytis retweeted
Come work with us! Opening for Assoc Teaching Prof, University of Cambridge. Expertise in any field of (or related to) machine learning/machine intelligence appropriate. Involves directing the MLMI MPhil programme and research e.g. building your own group.
jobs.cam.ac.uk/job/35215
Ignas Budvytis retweeted
Had a nice afternoon out in the sun with @robertocipolla and @IgnasBud, testing our 3D shape and pose estimation method 😁
Paper + code here: github.com/akashsengupta1997…