Agent skill · research science · aradotso
lingbot-map-3d-reconstruction
Feed-forward 3D foundation model for streaming scene reconstruction using Geometric Context Transformer
Why this skill is useful
Adds executable scripts for 3D scene reconstruction from images and video, leveraging a specialized model not commonly found in public resources.
What it needs
Requires conda, flashinfer-python, onnxruntime, onnxruntime-gpu, torch installed locally. About 6k tokens when loaded. Last updated 2026-07-09. 70 stars on the source repository.
What this skill does
LingBot-Map 3D Reconstruction Skill Skill by ara.so — Daily 2026 Skills collection. LingBot-Map is a feed-forward 3D foundation model that reconstructs scenes from streaming image or video data using a Geometric Context Transformer. It achieves ~20 FPS on 518×378 resolution over sequences exceeding 10,000 frames via paged KV cache attention. What It Does Streaming 3D reconstruction from image sequences or video Feed-forward inference (no iterative optimization needed) Outputs: point clouds with per-point confidence, camera poses, depth maps Key features: anchor context, pose-reference window, trajectory memory for drift correction Installation Model Download Models available on HuggingFace and ModelScope: Or manually download from: HuggingFace: https://huggingface.co/robbyant/lingbot-map ModelScope: https://www.modelscope.cn/models/Robbyant/lingbot-map CLI Commands Demo with Interactive 3D Viewer (browser at localhost:8080) Long Sequence Handling Without FlashInfer (SDPA fallback) Sky Masking with Custom Paths CLI Arguments Reference Input Argument Description :--- :--- --modelpath Path to model checkpoint (.pt file) --imagefolder Directory of input images --videopath Input video file path --fps Frames per second to sample from video Inference Mode Argument Default Description :--- :--- :--- --mode streaming streaming or windowed --windowsize 64 Window size for windowed mode --keyframeinterval 1 Store every Nth frame in KV cache --usesdpa False Use PyTorch SDPA instead of FlashInfer Sky Masking Argument Description :--- :--- --masksky Enable sky segmentation and masking --skymaskdir Custom directory for cached sky masks --skymaskvisualizationdir Save side-by-side mask visualizations Visualization Argument Default Description :--- :--- :--- --port 8080 Viser viewer port --confthreshold 1.5 Filter low-confidence points --pointsize 0.00001 Point cloud point size --downsamplefactor 10 Spatial downsampling for display Python API Usage Basic Streaming Inference Loading an …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills aradotso/lingbot-map-3d-reconstruction