VersaQ-3D: A Reconfigurable Accelerator Enabling Feed-Forward and Generalizable 3D Reconstruction via Versatile Quantization

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Zhang, Yipu, Cheng, Jintao, Liu, Xingyu, Li, Zeyu, Li, Carol Jingyi, Wu, Jin, Jiang, Lin, Xie, Yuan, Xu, Jiang, Zhang, Wei
Format:	Preprint
Published:	2026
Subjects:	Hardware Architecture
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

_version_	1866914286057553920
author	Zhang, Yipu Cheng, Jintao Liu, Xingyu Li, Zeyu Li, Carol Jingyi Wu, Jin Jiang, Lin Xie, Yuan Xu, Jiang Zhang, Wei
author_facet	Zhang, Yipu Cheng, Jintao Liu, Xingyu Li, Zeyu Li, Carol Jingyi Wu, Jin Jiang, Lin Xie, Yuan Xu, Jiang Zhang, Wei
contents	The Visual Geometry Grounded Transformer (VGGT) enables strong feed-forward 3D reconstruction without per-scene optimization. However, its billion-parameter scale creates high memory and compute demands, hindering on-device deployment. Existing LLM quantization methods fail on VGGT due to saturated activation channels and diverse 3D semantics, which cause unreliable calibration. Furthermore, VGGT presents hardware challenges regarding precision-sensitive nonlinear operators and memory-intensive global attention. To address this, we propose VersaQ-3D, an algorithm-architecture co-design framework. Algorithmically, we introduce the first calibration-free, scene-agnostic quantization for VGGT down to 4-bit, leveraging orthogonal transforms to decorrelate features and suppress outliers. Architecturally, we design a reconfigurable accelerator supporting BF16, INT8, and INT4. A unified systolic datapath handles both linear and nonlinear operators, reducing latency by 60%, while two-stage recomputation-based tiling alleviates memory pressure for long-sequence attention. Evaluations show VersaQ-3D preserves 98-99% accuracy at W4A8. At W4A4, it outperforms prior methods by 1.61x-2.39x across diverse scenes. The accelerator delivers 5.2x-10.8x speedup over edge GPUs with low power, enabling efficient instant 3D reconstruction.
format	Preprint
id	arxiv_https___arxiv_org_abs_2601_20317
institution	arXiv
publishDate	2026
record_format	arxiv
spellingShingle	VersaQ-3D: A Reconfigurable Accelerator Enabling Feed-Forward and Generalizable 3D Reconstruction via Versatile Quantization Zhang, Yipu Cheng, Jintao Liu, Xingyu Li, Zeyu Li, Carol Jingyi Wu, Jin Jiang, Lin Xie, Yuan Xu, Jiang Zhang, Wei Hardware Architecture The Visual Geometry Grounded Transformer (VGGT) enables strong feed-forward 3D reconstruction without per-scene optimization. However, its billion-parameter scale creates high memory and compute demands, hindering on-device deployment. Existing LLM quantization methods fail on VGGT due to saturated activation channels and diverse 3D semantics, which cause unreliable calibration. Furthermore, VGGT presents hardware challenges regarding precision-sensitive nonlinear operators and memory-intensive global attention. To address this, we propose VersaQ-3D, an algorithm-architecture co-design framework. Algorithmically, we introduce the first calibration-free, scene-agnostic quantization for VGGT down to 4-bit, leveraging orthogonal transforms to decorrelate features and suppress outliers. Architecturally, we design a reconfigurable accelerator supporting BF16, INT8, and INT4. A unified systolic datapath handles both linear and nonlinear operators, reducing latency by 60%, while two-stage recomputation-based tiling alleviates memory pressure for long-sequence attention. Evaluations show VersaQ-3D preserves 98-99% accuracy at W4A8. At W4A4, it outperforms prior methods by 1.61x-2.39x across diverse scenes. The accelerator delivers 5.2x-10.8x speedup over edge GPUs with low power, enabling efficient instant 3D reconstruction.
title	VersaQ-3D: A Reconfigurable Accelerator Enabling Feed-Forward and Generalizable 3D Reconstruction via Versatile Quantization
topic	Hardware Architecture
url	https://arxiv.org/abs/2601.20317