TrianguLang: Geometry-Aware Semantic Consensus for Pose-Free 3D Localization

Paper: arXiv:2603.08096 Project Page: cwru-aism.github.io/triangulang Code: github.com/bryceag11/triangulang

Bryce Grant, Aryeh Rothenberg, Atri Banerjee, Peng Wang — Case Western Reserve University

Overview

TrianguLang is a feed-forward, pose-free method for language-guided 3D localization from multi-view images. Given unposed images and a text query, it produces per-view segmentation masks and camera-relative 3D locations at ~10 FPS.

Checkpoints

Checkpoint	Description
mo_v11/best.pt	Multi-object (text + spatial), 230 scenes, 8 views, 100 epochs
fullscale_no_qp/best.pt	Single-object (text-only), 230 scenes, 100 epochs

Architecture

• Frozen: SAM3 (841M) + DA3-NESTED-GIANT-LARGE (1.69B) = ~2.5B params
• Trainable: GASA Decoder (~13.5M params)

Results (ScanNet++)

Setting	mIoU	mAcc
Text-only (single-object)	62.4%	77.4%
Text-only + CRF	65.2%	-

Citation

@article{grant2026triangulang,
  title={TrianguLang: Geometry-Aware Semantic Consensus for Pose-Free 3D Localization},
  author={Grant, Bryce and Rothenberg, Aryeh and Banerjee, Atri and Wang, Peng},
  journal={arXiv preprint arXiv:2603.08096},
  year={2026}
}