EgoFun3D: Modeling Interactive Objects from Egocentric Videos using Function Templates
Paper • 2604.11038 • Published
Acknowledge license to accept the repository
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
By using data from this repo, users agree to the corresponding licensing of the original data: - Ego-Exo4D license. - FunGraph3D license (CC BY-NC 4.0).
To request access to EgoFun3D, you will need to provide your full name (please provide both your first and last name), the name of your advisor or the principal investigator (PI) of your lab (in the PI/Advisor) fields, and the school or company that you are affiliated with (the Affiliation field). After requesting access to this repo, you will be considered for access approval.
Log in or Sign Up to review the conditions and access this dataset content.
This dataset contains egocentric videos and paired interactive objects as described in EgoFun3D: Modeling Interactive Objects from Egocentric Videos using Function Templates. Some raw data is sourced from Ego-Exo4D, FunGraph3D, and some self-captured data. We thank the authors for open-sourcing their datasets.
File Structure
Our dataset follows this file structure
full_dataset
|--articulation_annotation
|--{object_id}.articulations.json
|--cameras
|--{video_name}
|--extrinsics.npy
|--intrinsics.json
|--function_annotation
|--{object_id}_function.json
|--geometry
|--{object_id}.ply
|--masks_h5
|--{video_name}.h5
|--part_annotation
|--{object_id}.json
|--videos
|--{video_name}
|--undistorted.mkv
|--cropped.mkv
|--dataset_metadata.json
File Formats
Metadata
dataset_metadata.json contains one JSON blob for each video sequence, defining all related annotations. Please see one example below.
{
"video_name": "egoexo_82_2_0_indiana_cooking_10_2_keystep_492_Adjust-the-stove-heat_31",
"source": "egoexo4d",
"video_mask_path": "masks_h5/egoexo_82_2_0_indiana_cooking_10_2_keystep_492_Adjust-the-stove-heat_31.h5",
"camera_extrinsics_path": "cameras/egoexo_82_2_0_indiana_cooking_10_2_keystep_492_Adjust-the-stove-heat_31/extrinsics.npy",
"camera_intrinsics_path": "cameras/egoexo_82_2_0_indiana_cooking_10_2_keystep_492_Adjust-the-stove-heat_31/intrinsics.json",
"geometry_path": "geometry/egoexo_new_82_2.ply",
"geometry_type": "mesh",
"part_annotation_path": "part_annotation/egoexo_new_82_2.json",
"articulation_path": "articulation_annotation/egoexo_new_82_2.articulations.json",
"function_annotation_path": "function_annotation/egoexo_new_82_2_function.json",
"function_instance_id": "00",
"object_name": "stove",
"object_id": "egoexo_new_82_2",
"original_video_path": "videos/egoexo_82_2_0_indiana_cooking_10_2_keystep_492_Adjust-the-stove-heat_31/undistorted.mkv",
"cropped_video_path": "videos/egoexo_82_2_0_indiana_cooking_10_2_keystep_492_Adjust-the-stove-heat_31/cropped.mkv"
}
video_name: the global unique identifier for video sequences.source: the source of the video.original_video_path: the relative path to the original mkv video.cropped_video_path: the relative path to the cropped mkv video (only for videos from egoexo4d).video_mask_path: relative path to the mask file of the video.camera_extrinsics_path: relative path to the camera extrinsics file.camera_intrinsics_path: relative path to the camera intrinsics file.geometry_path: relative path to the geometry file.geometry_type: type of the geometry, can be "mesh" or "point cloud".part_annotation_path: relative path to the geometry part annotation file.articulation_path: relative path to the articulation annotation file.function_annotation_path: relative path to the part functionality annotation file.function_instance_id: the unique identifier for the part functionality for the corresponding geometry.object_name: name of the object.object_id: the global unique identifier for the object.
Articulation
Each object has an articulation annotation file, containing annotated parameters for every articulated joint of the object. One articulated joint annotation example is shown below.
[
{
"pid": 1,
"type": "revolute",
"base": [
0
],
"axis": [
-0.6887702788689056,
0.7249796569193167,
1.6097782149933394e-16
],
"origin": [
-1.99663739978191,
0.437388592032106,
-1.3537882566452
],
"rangeMin": 0,
"rangeMax": 0.40142572795869574
}
]
pid: part id of the child link on the mesh/point cloud annotation, corresponding to the part annotation file.type: joint type.base: parent link id.axis: joint axis.origin: joint position for revolute joint.rangeMin: minimum joint value.rangeMax: maximum joint value.
Camera
extrinsics.npy: $N\times 4\times 4$ camera to world matrix in opencv convention.Nis the number of frames.intrinsics.json:{ "undistorted_intrinsics": [ [ 608.2153458675755, 0.0, 703.5 ], [ 0.0, 608.2153458675755, 703.5 ], [ 0.0, 0.0, 1.0 ] ], "cropped_intrinsics": [ [ 608.2153458675755, 0.0, 497.5 ], [ 0.0, 608.2153458675755, 497.5 ], [ 0.0, 0.0, 1.0 ] ], "cropped_top_left": [ 206, 206 ], "cropped_bottom_right": [ 1200, 1200 ], "original_frame_size": [ 1408, 1408 ] }undistorted_intrinsics: intrinsic matrix for undistorted frames.cropped_intrinsics: intrinsic matrix for cropped frames.cropped_top_left,cropped_bottom_right: cropped uv coordinates on the undistorted frames.original_frame_size: $H\times W$ for the undistorted frames.
Function
Each object has a function annotation file, containing a list of annotated function templates for every part functionality of the object. One function template annotation example is shown
[
{
"physics": "geometry",
"func": "linear",
"function_instance_id": "00"
}
]
physics: physical effect.func: mapping.function_instance_id: function instance id for specific scene/object.
Geometry
It is a reconstruction of scene/object. Note that the ply file is either a point cloud or a mesh.
Mask
Stored in a HDF5 file with the following structure
{
{name}:
{
"mask_idx": 1,
"masks": masks
}
}
name: name for each mask.mask_idx: mask id.masks: $N\times H\times W$ numpy array.- 1 for left hand, 2 for right hand, 3 for receptor, 4 for effector, 5 for the whole object, 6(optional) for other important parts / objects.
Part
{
"01": {
"interection_id": 1,
"scene_id": "6",
"description": "control the water flow",
"receptor": {
"label": "faucet handle",
"indices": [
641,
642,
643,
644,
645,
...
],
"pid": 1
},
"effector": {
"label": "faucet spout",
"indices": [
2868,
2985,
2986,
2987,
3103,
...
],
"pid": 2
}
}
}
01: function instance id for specific scene/object.interaction_id: same for video clip id.scene_id: associated scene/object id.description: natural language description of the part functionality.label: part name.indices: vertex indices for the mesh or point indices for the point cloud for the part.pid: id for this part.
Video
undistorted.mkv: undistorted video.cropped.mkv: cropped video from the undistorted video to remove black areas.
Citation
If you find our dataset to be useful, please cite our paper
@article{peng2026egofun3d,
title={{EgoFun3D: Modeling Interactive Objects from Egocentric Videos using Function Templates}},
author={Peng, Weikun and Iliash, Denys and Savva, Manolis},
journal={arXiv preprint arXiv:2604.11038},
year={2026}
}
- Downloads last month
- 7