Kirill Mazur
@makezur
Followers
346
Following
867
Media
5
Statuses
18
PhD student at Dyson Robotics Lab @ Imperial College London
London, UK
Joined May 2020
SuperPrimitives enable 3D scene reconstruction at a level of local image regions rather than pixels. With SuperPrimitves, 3D reconstruction is done simply, thanks to powerful off-the-shelf 2D priors. project page: https://t.co/l360soWiAz video: https://t.co/H3GJ81qxx2
3
46
190
Introducing MASt3R-SLAM, the first real-time monocular dense SLAM with MASt3R as a foundation. Easy to use like DUSt3R/MASt3R, from an uncalibrated RGB video it recovers accurate, globally consistent poses & a dense map. With @eric_dexheimer*, @AjdDavison (*Equal Contribution)
42
260
1K
Demo tomorrow! https://t.co/DjF5ZgVvE5
Come by our poster this afternoon! ๐ญ๐ณ:๐ญ๐ฑ ๐๐ผ ๐ญ๐ด:๐ฐ๐ฑ, ๐๐ฟ๐ฐ๐ต ๐ฐ๐-๐
0
0
2
Curious how DSINE surface normals can be converted into full 3D geometry? Visit our demo of ๐ฆ๐๐ฝ๐ฒ๐ฟ๐ฃ๐ฟ๐ถ๐บ๐ถ๐๐ถ๐๐ฒ-based real-time 3D reconstruction tomorrow (Friday)! A demo is worth a thousand pictures ๐ Demo: 10:30 to 18:45, Arch CDE #9
#CVPR #CVPR2024
SuperPrimitives will be presented at #CVPR next week (Wednesday), along with a ๐ฟ๐ฒ๐ฎ๐น-๐๐ถ๐บ๐ฒ ๐ฑ๐ฒ๐บ๐ผ on Friday! Our new representation enables dense monocular 3D reconstruction in real-time. No poses required! Project page: https://t.co/l360soWiAz
0
1
14
Come by our poster this afternoon! ๐ญ๐ณ:๐ญ๐ฑ ๐๐ผ ๐ญ๐ด:๐ฐ๐ฑ, ๐๐ฟ๐ฐ๐ต ๐ฐ๐-๐
SuperPrimitives will be presented at #CVPR next week (Wednesday), along with a ๐ฟ๐ฒ๐ฎ๐น-๐๐ถ๐บ๐ฒ ๐ฑ๐ฒ๐บ๐ผ on Friday! Our new representation enables dense monocular 3D reconstruction in real-time. No poses required! Project page: https://t.co/l360soWiAz
3
9
51
SuperPrimitives will be presented at #CVPR next week (Wednesday), along with a ๐ฟ๐ฒ๐ฎ๐น-๐๐ถ๐บ๐ฒ ๐ฑ๐ฒ๐บ๐ผ on Friday! Our new representation enables dense monocular 3D reconstruction in real-time. No poses required! Project page: https://t.co/l360soWiAz
3
43
213
How can we infer 3D-consistent poses and dense geometry in real-time given only RGB images? ๐๐ข๐ ๐ข decodes dense geometry from a compact and optimizable set of 3D anchor points to enforce 3D consistency. Project page: https://t.co/57wOkz6McV Work with @AjdDavison 1/n
4
56
267
Code release for SuperPrimitives, and it comes with an interactive GUI! #CVPR2024
https://t.co/ZWGwYhbVCF SuperPrimitive is a new 3D representation which enables solving many 3D tasks at the level of image segments.
1
38
172
SuperPrimitives got accepted to #CVPR2024! The code will be released soon and see you all in Seattle!
SuperPrimitives enable 3D scene reconstruction at a level of local image regions rather than pixels. With SuperPrimitves, 3D reconstruction is done simply, thanks to powerful off-the-shelf 2D priors. project page: https://t.co/l360soWiAz video: https://t.co/H3GJ81qxx2
2
7
80
Survival of the fittest!
Excited to announce Fit-NGP which will be presented in #ICRA2024! Fit-NGP accurately estimates 6-DoF object poses (~ 1.6mm) leveraging Instant-NGP's density field. With @alzugarayign & @AjdDavison. Project page: https://t.co/RMMz4hKthj Video: https://t.co/5W9LRmJtxq (1/3)
0
0
8
Should the complexity of 3D reconstruction tasks be dependent on # of pixels or # of โthingsโ in the image? We use surface normal and segmentation priors to split the image into 2.5D segments and show how this representation can help tackle a wide range of 3D reconstruction tasks
SuperPrimitives enable 3D scene reconstruction at a level of local image regions rather than pixels. With SuperPrimitves, 3D reconstruction is done simply, thanks to powerful off-the-shelf 2D priors. project page: https://t.co/l360soWiAz video: https://t.co/H3GJ81qxx2
1
7
31
SuperPrimitives are constructed by splitting an image into a set of regions with a segmentation network. Think of it as a modern analogue of *superpixels*! To get its 3D shape, each segment is enhanced with surface normals, predicted by another network.
0
0
6
Our work on leveraging general pre-trained feature extractors for open-set real-time scene understanding has been accepted for #ICRA2023! Joint work with @SucarEdgar @AjdDavison Check out project page:
New: Feature-realistic neural fusion for real-time, open set scene understanding. Our neural field renders to feature space, enabling real-time grouping and segmentation of similar objects or parts from ultra-sparse, online interaction. Dyson Robotics Lab. https://t.co/vUn96ubJJe
0
6
44
Can you get SoTA results on 4 various point cloud processing tasks with a single block? The answer is yes! ๐Project page: https://t.co/9uHrNARe60 ๐ฝ๏ธVideo: https://t.co/EPSzcyw1po ๐งPaper: https://t.co/uTddO6cntc Join us on Session 8b at #ICCV21 on Oct 15th
2
1
23
Excited to share our work on Point-Based Modeling of Human Clothing accepted to ICCV'21! ๐ฅณ ๐ Project page: https://t.co/IL7M3TFiwa ๐ฌ Video: https://t.co/b8ZTkMiV0v ๐ Paper: https://t.co/gfQ0kg3cg8 Join us on Session 11 at https://t.co/P1jSQSjbi6 on Oct 13th and 15th
4
9
67