ARKitScenes
Indoor RGB-D captures with camera trajectories, reconstructed geometry, and scene annotations.
About
ARKitScenes connects mobile RGB video, depth, poses, and reconstructed indoor geometry. It is particularly useful when a pipeline needs to reason across the original observations and a shared 3D scene.
Use the interactive recording to inspect alignment between cameras, depth, meshes, and annotations. Small transform mistakes become obvious when every representation is visible together.
Suggested agent workflows
Copy a prompt into your agent of choice. Each workflow starts small, respects the source terms, and makes the result inspectable in Rerun.
Audit a training slice
Fetch ARKitScenes from its official source at https://github.com/apple/ARKitScenes. Read and preserve its source terms and citation requirements. Build a reproducible manifest for a representative slice, then load it into Rerun. Inspect the RGB video, Depth, Meshes, Poses streams for timestamp gaps, missing samples, calibration problems, and annotation inconsistencies. Summarize the findings and save both the validation script and a Rerun recording that makes every issue easy to review.
Build a baseline pipeline
Use ARKitScenes to scaffold a baseline pipeline for 3D reconstruction, Scene understanding, Depth. Start with a small deterministic cohort and document the selection criteria. Create the dataset adapter, preprocessing steps, train and validation splits, and one lightweight baseline. Log inputs, labels, predictions, losses, and representative failures to Rerun so I can inspect the full experiment timeline. Include exact setup and reproduction commands.
Find useful failure cohorts
Analyze ARKitScenes and propose a practical cohort taxonomy for 3D reconstruction, Scene understanding, Depth. Use available metadata and the RGB video, Depth, Meshes, Poses signals to identify hard cases, rare conditions, coverage gaps, and possible leakage between splits. Materialize the cohorts as queryable metadata, log examples from each cohort to Rerun, and produce a short report explaining which cohorts should become regression tests. Do not infer labels that are not present in the source data.