ASTRM

22 posts

ASTRM banner
ASTRM

ASTRM

@astrm_labs

ASTRM Intelligence Labs

Planet Earth Katılım Haziran 2025
31 Takip Edilen454 Takipçiler
ASTRM
ASTRM@astrm_labs·
sim2world2sim2world2sim2world2sim...
ud@uddupa

@sha0nk breathing life to sim room @astrm_labs with projection mapping tools built from scratch!

English
0
1
15
2.6K
ASTRM
ASTRM@astrm_labs·
🚀🚀
Shivam Kajale@shivam_kajale

Had a lovely sprint here at @astrm_labs over the last 36 hours. From unpacking servo motors, through designing CAD models for the turret frame, 3D printing the components, powering up the system, writing a firmware and driver, bringing it on the ASTRM operating system, to getting it to following targets using a vision pipeline - loved every moment of it. Still a long journey ahead, but off to an invigorating start!

ART
0
0
4
590
ASTRM
ASTRM@astrm_labs·
The draft DAP 2026 is a massive win for true deep-tech. By shifting the focus from manufacturing localization to IP ownership and Design Sovereignty, the @DefenceMinIndia is clearing the path for companies like ours that prioritize original engineering over licensed production. We are ready to help build India into a global design powerhouse! 🇮🇳
Shivam Kajale@shivam_kajale

The draft Defence Aquisition Procedure 2026 just dropped and it feels like a gift! 🚀 I started @astrm_labs with this year, with my friend @uddupa , with a singular focus: doing the heavy lifting on IP creation for emerging defense tech. I took this plunge on the conviction that what the Indian defense industry needs isn't more licensed assembly lines, but true Design Authority. We’re realizing this by building a high-caliber team of the finest engineers and designers - people who want to own the source code and the core architecture, not just follow a foreign blueprint. The draft DAP2026 is a strong validation for our thesis. It totally shows a doctrinal shift on the MoD’s side - moving its primary metric of success from "Made in India" (manufacturing localization) to "Owned by India" (IP ownership and design sovereignty). The pivot is clear - DAP 2020: Indigenization. It rewarded you for manufacturing a foreign gun using Indian steel and labor. DAP 2026: Sovereignty. It rewards you if you own the Source Code, the Gerber Files, and the Algorithms. It explicitly separates "License Production" from true indigenization. Huge win for deep-tech. 🇮🇳 Will follow up with a thread on my biggest takeaways for deep-tech startups building 'Owned by India' IP.

English
0
0
3
1.3K
ASTRM retweetledi
Robots Digest 🤖
Robots Digest 🤖@robotsdigest·
Physical intelligence needs touch, not just vision. eFlesh is a fully 3D-printable magnetic tactile sensor that enables reliable slip detection and contact awareness on real robots, using low-cost hardware and scalable fabrication
English
16
211
2K
243K
ASTRM retweetledi
sasaki@engineer
sasaki@engineer@rsasaki0109·
Enhanced SLAM3R: Real-Time Reconstruction via Online Camera Stream github.com/HJCheng0602/SL… This is an enhanced and improved version of the well-known SLAM3R framework, designed to significantly expand its input versatility. The core advancement lies in enabling the use of online camera feeds, empowering users to perform remote and real-time 3D reconstruction using readily available mobile device cameras, thus overcoming the limitations of static local video files.
English
0
21
136
8K
ASTRM retweetledi
Kwang Moo Yi
Kwang Moo Yi@kwangmoo_yi·
Zhao and Wei et al., "Spatia: Video Generation with Updatable Spatial Memory" "Memory" for video models via point cloud-conditioned video generation. I am obviously still biased towards having these "explicit" 3D stuff.
English
8
26
239
14.4K
ASTRM retweetledi
Basile Terver @ ICLR 2026
Basile Terver @ ICLR 2026@BasileTerv987·
My first PhD paper is out! 🎓 "What Drives Success in Physical Planning with Joint-Embedding Predictive World Models?" tl:dr: JEPA-WMs for robotics: learn dynamics on top of visual encoders, optimize actions towards goal 👇 w/ @JimmyTYYang1, Jean Ponce, @AdrienBardes, @ylecun
English
13
109
916
80.4K
ASTRM retweetledi
sasaki@engineer
sasaki@engineer@rsasaki0109·
InfiniteVGGT: Visual Geometry Grounded Transformer for Endless Streams Achieving higher reconstruction quality and more accurate camera pose estimation using thousands of frames input. github.com/AutoLab-SAI-SJ… >We propose InfiniteVGGT, a causal visual geometry transformer that utilizes a training-free rolling memory mechanism to enable stable, infinite-horizon streaming, and introduce the Long3D benchmark to rigorously evaluate long-term continuous 3D geometry performance. Our main contributions are summarized as follows: 1. An unbounded memory architecture \mymethod{} for continuous 3D geometry understanding, built on a novel, dynamic, and interpretable explicit memory system. 2. State-of-the-art performance on long-sequence benchmarks and a unique capability for robust, infinite-horizon reconstruction without memory overflow. 3. The Long3D benchmark, a new dataset for the rigorous evaluation of long-term performance, addressing a critical gap in the field.
sasaki@engineer tweet media
English
1
11
93
5.3K
ASTRM retweetledi
sasaki@engineer
sasaki@engineer@rsasaki0109·
Pi3MOS-SLAM Dynamic Visual SLAM using a General 3D Prior github.com/PRBonn/Pi3MOS-… Reliable incremental estimation of camera poses and 3D reconstruction is key to enable various applications including robotics, interactive visualization, and augmented reality. However, this task is particularly challenging in dynamic natural environments, where scene dynamics can severely deteriorate camera pose estimation accuracy. In this work, we propose a novel monocular visual SLAM system that can robustly estimate camera poses in dynamic scenes. To this end, we leverage the complementary strengths of geometric patch-based online bundle adjustment and recent feed-forward reconstruction models. Specifically, we propose a feed-forward reconstruction model to precisely filter out dynamic regions, while also utilizing its depth prediction to enhance the robustness of the patch-based visual SLAM. By aligning depth prediction with estimated patches from bundle adjustment, we robustly handle the inherent scale ambiguities of the batch-wise application of the feed-forward reconstruction model.
sasaki@engineer tweet media
English
0
17
136
6.9K
ASTRM retweetledi
Thomas Richter-Trummer
Thomas Richter-Trummer@PixelBilly·
Capturing Unreal scenes with 360° equirectangular images → fixed-pose COLMAP rig → training becomes much more predictable for #gaussiansplatting. Unreal Capture → COLMAP → Lichtfeld Studio (MCMC Default) → LOD (SOGv2) + FX→ runs on the web and mobile. Demo link in reply
English
2
30
242
27.3K