Alexreysa/wide-robot-object-inside-container-v0

1---2pretty_name: wide-robot object_inside_container (real-camera v0)3task_categories:4- robotics5tags:6- robotics7- manipulation8- object-inside-container9- apriltag10- fiducial-markers11- verifier12- multi-view13size_categories:14- n<1K15license: other16license_name: tbd17---18 19# wide-robot — `object_inside_container` real-camera capture (v0)20 21Raw multi-view video for the **Phase 3A real-camera ingestion** pilot of22[wide-robot](https://github.com/alex-reysa/wide-robot): a marker-based, **source-independent23verifier** judges `object_inside_container` episodes (PASS / FAIL / UNCERTAIN) from real camera24evidence. This repo holds the **raw capture media** that is gitignored in the code repo; the derived25JSON (calibrations, tracks, rollouts, verdicts) and the full results live in the git repo under26`datasets/sony_object_inside_container_v0/` (see `INGESTION_RESULTS.md`).27 28> **Conservative pilot, not a benchmark.** The associated result is deliberately narrow: across 7829> clips the frozen verifier produced **0 false PASSes on 30 genuine-failure clips**, with partial30> success recall — a demonstration that the verifier's *safety* survives real, approximate-calibration31> evidence. It is marker-based and calibration-limited, not a high-recall perception system.32 33## Contents34 35- **80 clips** = 40 episodes × 2 simultaneous cameras (`sony_front` ≈ 45° frontal; `iphone_top` top36  view), under `raw_videos/<category>/oic_<category>_<seq:03d>__<camera>.mp4`.37- `manifest.json` — provenance index: per-clip SHA-256, expected class, camera, video metadata, and38  the as-built capture metadata (this is the source of truth).39- `labels.csv` — per-clip expected verdicts for the two bundled targets.40- `SHA256SUMS` — independent checksums over every `.mp4` (verify with `shasum -a 256 -c SHA256SUMS`).41 42### Categories43`success`, `fail_near_not_inside`, `fail_on_rim`, `fail_outside`, `control_static`,44`control_born_inside`, `control_born_inside_hand_motion`, `control_inside_to_inside`,45`control_inside_to_outside`, `success_tag_obstruction`, `success_hand_obstruction`, `calibration`.46 47## Capture setup (as-built; `manifest.json` is authoritative)48 49- Cube `5×5×5 cm`; AprilTag (`tag36h11`) IDs **2** (top) / **3** (front), 35 mm.50- Tray ~`18×18 cm`, ~4 cm rim — **homemade cardboard, not perfectly square/consistent**; ingestion51  fits the geometry. Tray tags **6** (front, outer) / **7** (inside floor), 50 mm.52- Table tags **0 / 1** (75 mm) — used only to recover the world frame. Tags 4/5 absent.53 54## Using it55 56```bash57hf download Alexreysa/wide-robot-object-inside-container-v0 --repo-type dataset --local-dir recordings58shasum -a 256 -c recordings/SHA256SUMS          # verify integrity59# then, from the wide-robot repo (pip install -e ".[camera]"):60python -m scripts.ingest_recordings --select all61```62 63## Provenance & licensing64 65- Source: tripod Sony A7IV + iPhone capture for `sony_object_inside_container_v0`; reproduced and66  judged by the byte-frozen verifier in the wide-robot repo. Every clip's SHA-256 is bound in67  `manifest.json` and `SHA256SUMS`.68- **License: TBD** — set by the dataset owner. Recommend CC-BY-4.0 (attribution) for reuse; until set,69  treat as all-rights-reserved.70