@marwan_ptr profile picture
Marwan Taher @marwan_ptr

PhD student at the Dyson Robotics Lab @ Imperial College London.

Joined April 2011
  • Tweets 75
  • Following 492
  • Followers 466
10 Photos and videos
Pinned Tweet
Marwan Taher@marwan_ptr
19 Dec 2025
How can we run reconstruction models like π³ and Depth Anything 3 in real-time? We present KV-Tracker, a training-free approach, for real-time tracking of scenes and objects. Achieving up to 30 FPS! With @alzugarayign, @makezur, @XinKong_IC and @AjdDavison
0:18
10
96
704
65,397
Wanna see a real-time demo of KV-Tracker? Stop by in ExHall F today 11:45 - 13:45! #CVPR2026
0:24
2
12
55
4,739
Project page: marwan99.github.io/kv_tracke… Code: github.com/Marwan99/kv_track…

KV-Tracker: Real-Time Pose Tracking with Transformers

Real-time 6-DoF pose tracking by caching key-value pairs from multi-view geometry networks. Up to 15× speedup at ~30 FPS without drift or catastrophic forgetting.

marwan99.github.io
1
1
203
Marwan Taher retweeted
Kirill Mazur@makezur
Jun 5
How to build complete 4D reconstructions from videos? Swing by 4DPM oral presentation today: Mile High Ballroom 3A - 4A, 13:00 #CVPR2026
0:07
3
8
108
10,408
Marwan Taher retweeted
Kirill Mazur@makezur
Apr 16
4DPM got selected for an oral presentation at #CVPR2026 🤠
Kirill Mazur@makezur
18 Dec 2025
Introducing 4D Primitive-Mâché (4DPM), a new method for replayable 4D reconstruction from monocular videos. We split dynamic scenes into 3D primitives and recover their motion. 4DPM can infer object positions even after they leave view. Joint work with @marwan_ptr @AjdDavison
0:30
3
5
65
6,956
KV-Tracker has been accepted to #CVPR2026!
Marwan Taher@marwan_ptr
19 Dec 2025
How can we run reconstruction models like π³ and Depth Anything 3 in real-time? We present KV-Tracker, a training-free approach, for real-time tracking of scenes and objects. Achieving up to 30 FPS! With @alzugarayign, @makezur, @XinKong_IC and @AjdDavison
0:18
2
19
224
16,885
How can we run reconstruction models like π³ and Depth Anything 3 in real-time? We present KV-Tracker, a training-free approach, for real-time tracking of scenes and objects. Achieving up to 30 FPS! With @alzugarayign, @makezur, @XinKong_IC and @AjdDavison
0:18
10
96
704
65,397
more replies
KV-Tracker enables object-level reconstruction and tracking when provided with an object mask. The KV-cache can be saved and used later without any special initialisation procedure.
0:34
2
19
2,048
More results and the paper can be found here: marwan99.github.io/kv_tracke… Video: youtu.be/ZVNnvZZxhoI

KV-Tracker: Real-Time Pose Tracking with Transformers

Real-time 6-DoF pose tracking by caching key-value pairs from multi-view geometry networks. Up to 15× speedup at ~30 FPS without drift or catastrophic forgetting.

marwan99.github.io
1
29
3,159
Per-frame geometry from π³ is split into primitives via segmentation and tracked over time using dense 2D point tracks. With a compact per primitive pose, geometry is densely aligned, stitching primitives to create a complete reconstruction of the observed scene components.
Kirill Mazur@makezur
18 Dec 2025
Introducing 4D Primitive-Mâché (4DPM), a new method for replayable 4D reconstruction from monocular videos. We split dynamic scenes into 3D primitives and recover their motion. 4DPM can infer object positions even after they leave view. Joint work with @marwan_ptr @AjdDavison
0:30
1
9
573
ACE-SLAM naturally handles loop closure without special treatment, robustly deals with dynamic objects, while remaining lightweight (small MLP) and computationally efficient—making this representation compelling for SLAM.
Ignacio Alzugaray@alzugarayign
17 Dec 2025
Excited to present ACE-SLAM, the first neural SLAM to use Scene Coordinate Regression as an implicit map representation Efficient (real-time from live stream), compressive (neural maps <1MB) and robust to dynamic scenes With @marwan_ptr and @AjdDavison ialzugaray.github.io/ace-sla…
0:18
2
10
999
Marwan Taher retweeted
Xin Kong@XinKong_IC
9 Sep 2025
🚀 Excited to share CausNVS: Autoregressive Multi-view Diffusion for Flexible 3D Novel View Synthesis! Let’s recon 3D world generatively. CausNVS handles any number of input views, synthesizes novel views autoregressively, enables interactive streaming and flexible N-to-M NVS.
0:07
2
19
105
10,278
Marwan Taher retweeted
Riku Murai@rmurai0610
16 Dec 2024
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)
0:34
42
253
1,433
203,567
EscherNet will be presented tomorrow at #CVPR. But *now* you can drop a couple of images into our Hugging Face demo to try it out! huggingface.co/spaces/kxic/E…

EscherNet - a Hugging Face Space by kxic

3D novel view synthesis from any number images!

huggingface.co
Xin Kong@XinKong_IC
19 Jun 2024
Tired of single image to 3D? Check out EscherNet tomorrow @CVPR that can take flexible number of views for 3D generation! THURSDAY, JUNE 20 ORAL: 9:00-10:30, SUMMIT BALLROOM (TOP FLOOR) POSTER: 10:30-12:00, ARCH 4A-E, #69 Try our @Gradio online demo huggingface.co/spaces/kxic/E…
0:56
6
428
Don't miss the real-time demo of SuperPrimitives at #CVPR!!
Kirill Mazur@makezur
14 Jun 2024
SuperPrimitives will be presented at #CVPR next week (Wednesday), along with a 𝗿𝗲𝗮𝗹-𝘁𝗶𝗺𝗲 𝗱𝗲𝗺𝗼 on Friday! Our new representation enables dense monocular 3D reconstruction in real-time. No poses required! Project page: makezur.github.io/SuperPrimi…
0:44
5
385
Marwan Taher retweeted
Aalok Patwardhan@AalokPat
26 Mar 2024
From RGB images we can estimate camera rotation, *without* knowledge of camera intrinsics. This also leads to some cool downstream applications - it can complement an IMU .. we call it U-ARE-ME! @AjdDavison @DoC_Rhodes94 @BaeGwangbin
Gwangbin Bae@BaeGwangbin
26 Mar 2024
𝗜𝗠𝗨? How about 𝗨-𝗔𝗥𝗘-𝗠𝗘? In this work, we show how monocular surface normal cues can be used for rotation estimation. callum-rhodes.github.io/U-AR… collab w/ @AalokPat, Callum Rhodes, @AjdDavison
0:19
1
6
41
4,419
Super impressive reconstruction quality!
Kirill Mazur@makezur
6 Mar 2024
SuperPrimitives got accepted to #CVPR2024! The code will be released soon and see you all in Seattle!
1
3
265
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: marwan99.github.io/Fit-NGP Video: youtu.be/KQ7yH_em3Qg (1/3)
0:13
4
34
181
35,358
Using *RGB* images, a NeRF is trained via Instant-NGP. A depth map of the object is rendered to obtain an initial coarse position estimate. The object model is then fitted to the reconstructed density field via a multi-hypothesis iterative optimization scheme. (2/3)
0:16
1
10
1,033
This allows highly accurate pose estimation even for challenging small and specular objects, which can enable precise manipulation. @ieee_ras_icra (3/3)
4
718
Load more