QVGGT：训练后量化的视觉几何基础 Transformer
QVGGT: Post-Training Quantized Visual Geometry Grounded Transformer
对需要在边缘设备部署前馈式三维重建模型的研究者，本文给出了按 Transformer 块敏感度分配精度的混合精度量化方案，以及用 PCA 全局补偿词元处理相机/寄存器离群词元的校准策略，可直接迁移到其他大规模几何 Transformer 的压缩。
原文依据 we introduce QVGGT, a tailored quantization framework designed to compress VGGT
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论随着视觉几何基础 Transformer（VGGT）的出现，直接从图像估计三维属性的能力进展迅速，该模型可在单次前向传播中预测相机参数、深度图与点云。然而，其 12 亿参数的规模严重限制了在无人机、移动 AR 设备等资源受限平台上的部署。为解决这一限制，我们提出 QVGGT，一个为压缩 VGGT 而设计的定制量化框架。我们的方法始于如下观察：VGGT 内部的 Transformer 块对量化表现出异质敏感性。因此，我们分析逐块量化敏感度，并提出一种选择性混合精度策略，为最脆弱的 Transformer 块分配更高精度。为应对高方差相机词元与寄存器词元所导致的量化误差放大，我们进一步引入带相机信息补偿的词元过滤，从激活校准中剔除这些离群值，并利用由 PCA 导出的全局补偿词元恢复其几何线索。最后，我们开发了一种任务感知的尺度搜索机制，不仅通过层重建，还通过多头监督以及相机位姿、深度图与点图之间的跨头几何一致性来评估候选量化尺度。在多个几何感知基准上的大量实验表明，QVGGT 实现了近乎无损的 W4A16 量化，在保持所有三维预测头精度的同时，相较 FP32 带来 3–4.9 倍内存降低与最高 2.8 倍的真实硬件加速。我们的方法使高保真三维感知在边缘设备上成为可能，从而推动前馈式三维重建模型在真实受限环境中的实际部署。
英文原文摘要 1781 字符
Estimating 3D attributes directly from images has advanced rapidly with the Visual Geometry Grounded Transformer (VGGT), which predicts camera parameters, depth maps, and point clouds in a single forward pass. However, its 1.2B-parameter scale severely limits deployment on resource-constrained platforms such as UAVs and mobile AR devices. To address this limitation, we introduce QVGGT, a tailored quantization framework designed to compress VGGT. Our approach starts from the observation that transformer blocks within VGGT exhibit heterogeneous sensitivity to quantization. We thus analyze per-block quantization sensitivity and propose a selective mixed-precision strategy that allocates higher precision to the most fragile transformer blocks. To address the amplification of quantization error caused by high-variance camera and register tokens, we further introduce token filtering with camera information compensation, which removes these outliers from activation calibration and restores their geometric cues using a PCA-derived global compensation token. Finally, we develop a task-aware scale search mechanism that evaluates candidate quantization scales not only through layer reconstruction but also through multi-head supervision and cross-head geometric consistency among camera poses, depth maps, and point maps.Extensive experiments on multiple geometry perception benchmarks demonstrate that QVGGT achieves near-lossless W4A16 quantization, preserving the accuracy of all 3D prediction heads while delivering 3 4.9xmemory reduction and up to 2.8x real hardware speedup over FP32.Our approach makes high-fidelity 3D perception feasible on edge devices, enabling practical deployment of feed-forward 3D reconstruction models in real-world constrained environments.