Datasets:

---

ACERobotics

/

ACE-Data-0

Like 58

Follow

ACE Robotics 38

Dear ACE-Data-0 Team,

Thank you for releasing this valuable dataset!

I am an undergraduate student working on a non-commercial academic research project involving real-to-simulation reconstruction and human-scene interaction. I am currently using ACE-Data-0 to build a physically interactive room in LiteReality, focusing on activities such as sitting on a sofa, picking up a cup, and using a water dispenser.

I have been using the released camera calibration, object motion data, and multi-view images to reconstruct the Room-scale environment. However, accurately determining the real-world dimensions, positions, and orientations of static furniture (such as the sofa, coffee table, bed, dining table, and kitchen furniture) remains challenging.

While examining the official visualization code (Room-scale/calibration/scripts/visualize/visualize_case.py), I noticed the following internal asset path:

/home/acerobotics/Datasets/mesh_alignmend/aligned_textured/

The code appears to support individual room object meshes (mesh.glb / mesh.obj) and room-level coordinate alignment. The documentation also mentions scene.obj, but indicates that optional room meshes are not included in the released calibration package.

I would be very grateful if you could clarify:

1. Room meshes: Are the reconstructed room meshes, including aligned_textured/ and possibly scene.obj, available to researchers through a separate download or access request?

2. Furniture geometry and layout: Do you have any metric furniture dimensions, instance positions/orientations, bounding boxes, or scene configuration files describing the actual Room-scale environment?

3. Coordinate alignment: If these assets are available, is there a final transformation aligning the room reconstruction with the OptiTrack Motive world coordinate system, including its units and scale?

4. Consistency across recordings: Does the Room-scale environment maintain the same furniture layout across takes, or are there session-specific scene configurations?

Even if the complete textured meshes cannot be shared, simplified collision geometry, a metric floor plan, or furniture poses and dimensions would be extremely helpful for our research.

Our goal is to reproduce the original physical environment as faithfully as possible, rather than reconstructing all furniture from scratch. I would be happy to follow any additional academic research access requirements.

Thank you very much for your time and for making this dataset available!

Best regards