--- pretty_name: AFUN license: cc-by-nc-sa-4.0 language: - en size_categories: - 10K//// # one folder per data point ├── obs_frame.png # RGB frame ├── obs_frame_depth.npy # float32 H×W depth, millimeters ├── sam_mask.png # affordance mask (non-zero = actionable region) └── trajectory.json # 3D motion + camera intrinsics ``` ## Load a data point ```python import json, numpy as np from PIL import Image m = json.load(open("manifest.json")) s = m["samples"][0] rgb = np.array(Image.open(f"{s['path']}/obs_frame.png")) depth = np.load(f"{s['path']}/obs_frame_depth.npy") # millimeters mask = np.array(Image.open(f"{s['path']}/sam_mask.png")) > 0 traj = json.load(open(f"{s['path']}/trajectory.json")) print(s["language"], traj["camera_info"]["intrinsics"]) ``` Each manifest entry has `path` (the folder), `dataset` / `episode_id` / `interval` / `cam`, the instruction (`language`, with variants in `queries`), and `shard` (which archive contains it). ## trajectory.json 3D positions are in the **camera frame, in meters**. `camera_info` holds the intrinsics (`fx, fy, cx, cy`), the distortion model, and `T_base_to_cam`. There are **two schemas**, because SceneFun3D scenes are annotated differently from robot and human videos: **A. Robot / human sources** (`droid`, `robomind`, `agibot`, `rh20t`, `rh20t_human`, `calvin`, `rlbench`, `vitra`) — one interaction per file, fields at the top level: | field | meaning | |---|---| | `trajectory_3d` | GT motion of the interaction point, `[{frame_idx, position_3d}, ...]` — the curve-fitted (denoised) track | | `motion_2d` | start / end pixel of the motion | | `spline_params.ctrl` | control points of the fitted 3D curve (the training target is sampled from this curve) | **B. `scenefun3d`** — SceneFun3D is a set of annotated 3D scans, not videos. A scene can have several annotated functional parts (a drawer, a window, a tap), so the motions live in a **list** called `trajectories`, one entry per annotation. Each entry has the same fields as schema A (`trajectory_3d`, `motion_2d`, `spline_params`) plus: | field | meaning | |---|---| | `annot_id` | the original SceneFun3D annotation id | | `interval_language` | the instruction for *this* annotation | | `motion_type` | `rot` (hinged: door, window) or `trans` (sliding: drawer) | | `scenefun3d_motion_params` | the analytic motion (see below) | `scenefun3d_motion_params` is what makes this source distinctive — the motion is given in closed form, not just as samples: - `motion_type: "rot"` → `motion_dir_cam` (rotation axis), `origin_cam` (a point on the axis, i.e. the hinge), `ref_cam` (reference point), `angle_rad` (e.g. `1.5708` = 90°) - `motion_type: "trans"` → `motion_dir_cam` (slide direction), `origin_cam` (start point), `distance_m` (e.g. `0.3`), `orient` (`inwards` / `outwards`) `trajectory_3d` is 15 points sampled from those parameters. **In practice `trajectories` has length 1** (~95% of files; the rest have 2). Reading either schema: ```python traj = json.load(open(f"{s['path']}/trajectory.json")) if "trajectories" in traj: # scenefun3d motion = traj["trajectories"][0] # [0] is enough for almost every file else: # robot / human sources motion = traj points = [p["position_3d"] for p in motion["trajectory_3d"]] # camera frame, meters ``` ## Sources | key | dataset | entries | |---|---|---:| | scenefun3d | SceneFun3D | 49,706 | | robomind | RoboMIND | 4,491 | | vitra | VITRA (human videos) | 2,432 | | droid | DROID | 1,664 | | rh20t_human | RH20T human demos | 644 | | agibot | AgiBot World | 614 | | rh20t | RH20T | 186 | | calvin | CALVIN | 118 | | rlbench | RLBench | 12 | | **total** | | **59,867** | The evaluation sets are released separately as [**AFUN_eval**](https://huggingface.co/datasets/afun-anonymous/AFUN_eval) and are disjoint from the training views at the sample level. `extra/` contains the test and validation views. ## Full index (59,867 entries) `manifest_full.json` lists every row of the index reported in the paper. Each entry has a `split`: | entries | count | archive | |---|---:|---| | training views | 44,749 | `data/*.tar.zst` (the same views as `manifest.json`) | | validation views (SceneFun3D) | 9,610 | `extra/val-scenefun3d-*.tar.zst` | | test views | 121 | `extra/test-0000.tar.zst` (also in AFUN_eval) | | views removed in later quality checks | 326 | `extra/removed-0000.tar.zst` | | repeated rows | 5,061 | `extra/duplicates-0000.tar.zst` | A repeated row lists a training view (4,977) or a test view (84) a second time. These entries extract to `duplicates/` and name the original in `duplicate_of`. There are 54,806 distinct views. Train on `manifest.json` only: `extra/` holds held-out data. Ego4D views in `extra/` are annotation-only (`sam_mask.png`, `trajectory.json`). ```bash for f in extra/*.tar.zst; do tar --zstd -xf "$f"; done ```