map anything

We have hosted the application map anything in order to run this application in our online workstations with Wine or directly.

Quick description about map anything:

Map-Anything is a universal, feed-forward transformer for metric 3D reconstruction that predicts a scene�s geometry and camera parameters directly from visual inputs. Instead of stitching together many task-specific models, it uses a single architecture that supports a wide range of 3D tasks�multi-image structure-from-motion, multi-view stereo, monocular metric depth, registration, depth completion, and more. The model flexibly accepts different input combinations (images, intrinsics, poses, sparse or dense depth) and produces a rich set of outputs including per-pixel 3D points, camera intrinsics, camera poses, ray directions, confidence maps, and validity masks. Its inference path is fully feed-forward with optional mixed-precision and memory-efficient modes, making it practical to scale to long image sequences while keeping latency predictable.

Features:

• One feed-forward transformer that covers >10 reconstruction tasks
• Multi-modal inputs (images, calibration, poses, depth) with unified APIs
• Dense metric outputs: 3D points, depth (z and along-ray), intrinsics, poses, ray directions, confidence and masks
• Turnkey demos plus exporters to COLMAP and Gaussian splatting pipelines
• Mixed-precision and memory-efficient inference for long sequences
• Modular �building blocks� (UniCeption, WAI) to scale data and models

Programming Language: Python.
Categories: