TL;DR FoundationStereo, the high-accuracy zero-shot stereo model, sped up by over 10x with almost no accuracy loss. It finally runs in real time. Title: Fast-FoundationStereo: Real-Time Zero-Shot Stereo Matching URL: Points 🎯 Problem: stereo had split into "accurate but heavy foundation models" vs "fast methods that sacrifice accuracy and generalization" 🧩 Divide-and-conquer: each of three bottlenecks accelerated separately (feature extraction, cost filtering, refinement) 📚 Feature extraction: distill hybrid monocular + stereo priors into a single efficient student 🔍 Cost filtering: blockwise NAS over 8 blocks cuts search from O(n^N) to O(n), with ILP minimizing error under a latency budget ✂️ Refinement: structured pruning via ConvGRU recurrent dependency graphs and Taylor-expansion importance 🗂 1.4M in-the-wild stereo pairs auto pseudo-labeled with a normal-consistency check ⚡ Middlebury-Q: 4.51% BP-2 at 49ms (vs 2.64% at 496ms, 10.1x slower), down to 21ms with TensorRT A new SOTA among real-time stereo methods, well suited for robotics and AR deployment. #StereoMatching# #ComputerVision#

Stereo depth models usually force a tradeoff. 📌 Paper + Code Either strong zero-shot generalization or real-time speed. Fast-FoundationStereo closes that gap. The model accelerates FoundationStereo by more than 10× while keeping comparable depth quality, making real-time stereo matching practical for robotics and XR. On an RTX 3090, it produces dense metric depth maps and 3D point clouds in real time, outperforming existing real-time stereo approaches and even challenging some heavier zero-shot models. The key idea is a divide-and-conquer architecture that removes bottlenecks in common stereo matching components while preserving the strengths of the original teacher model. Training also tackles a major constraint in stereo vision: the lack of ground-truth metric depth data. The team built an automatic data curation pipeline that generates pseudo-labels from internet-scale stereo imagery using the Stereo4D dataset. Result: a real-time stereo foundation model with strong zero-shot performance and practical runtime speed. Code, pretrained weights, and pseudo-labeled stereo data will be released. Thanks for sharing, @bowenwen_me! 📍Paper: Code: Project: ——- Weekly robotics and AI insights. Subscribe free: