Skip to content
RobotWorld
← Back to Videos

SpatialLM: Labeled 3D Room Layouts from Monocular Phone Video

Loading video
Loading video

SpatialLM is a 0.5B multimodal LLM that turns monocular RGB video into structured 3D room layouts: MASt3R-SLAM point clouds in, walls, doors, windows and oriented object boxes out. Trained as a standard multimodal LLM on 12,328 synthetic scenes, it reaches state-of-the-art layout estimation.

Spatial Intelligence3D Scene UnderstandingVision-language modelsScene ReconstructionManycore
Category: perception
Author: @0x0SojalSec
Date: 2026-10-09T00:00:00
Duration: 50.666s
Source: https://x.com/0x0SojalSec/status/2108683699130970241
Reference: https://github.com/manycore-research/SpatialLM

Related Videos

MultiSet AI Releases VPS Gen2: Camera-to-3D Map Alignment
0:45

MultiSet AI Releases VPS Gen2: Camera-to-3D Map Alignment

@tokufxug · 392 views · 2026-07-22
MultiSet AIVPS Gen2Gaussian Splat
ABot-Recon Streaming 3D Reconstruction
0:15

ABot-Recon Streaming 3D Reconstruction

@IndieDevHailey · 545 views · 2026-08-29
SLAMAMapABot-Recon
NVIDIA ArtiFixer Repairs Incomplete Gaussian Splats
0:29

NVIDIA ArtiFixer Repairs Incomplete Gaussian Splats

@justinryanio · 371 views · 2026-07-20
3D Gaussian SplattingNVIDIAGaussian Splatting
OneCanvas: reproject video frames onto a single panoramic canvas so a stock 2D VLM reasons in 3D — SOTA on SQA3D, VSI-Bench and SPBench at an order of magnitude less compute

OneCanvas: reproject video frames onto a single panoramic canvas so a stock 2D VLM reasons in 3D — SOTA on SQA3D, VSI-Bench and SPBench at an order of magnitude less compute

@unknown · 364 views · 2026-06-19
VLMSpatial reasoningOneCanvas
R3: streaming feed-forward 3D reconstruction via relative-pose regression — 372M params, 20+ FPS, thousands of frames, open-sourced for NeurIPS 2026
0:25

R3: streaming feed-forward 3D reconstruction via relative-pose regression — 372M params, 20+ FPS, thousands of frames, open-sourced for NeurIPS 2026

@CongrongX · 471 views · 2026-10-01
R33D reconstructionStreaming
Fish Audio upgrades its ASR model: speaker identification and inline emotion cues such as [laughter]
0:59

Fish Audio upgrades its ASR model: speaker identification and inline emotion cues such as [laughter]

@FishAudio · 674 views · 2026-09-28
ASRSpeech RecognitionEmotion Recognition

As an Amazon Associate, we earn from qualifying purchases.