Download

Dataset Overview

The Omni6DPose dataset is available for download at Omni6DPose. The dataset is organized into four parts:

ROPE: the real dataset for evaluation.
SOPE: the simulated dataset for training.
PAM: the pose aligned 3D models used in both ROPE and SOPE.
Meta: the meta information of the objects in PAM.

Omni6DPose provides a large-scale dataset with comprehensive data modalities and accurate annotations. The dataset is organized as follows. The files marked as Optional may not be necessary for some methods or tasks, but they are included to support a wide range of research possibilities.

Omni6DPose
├── ROPE
│   ├── SCENE_ID
│   │   ├── FRAME_ID_meta.json
│   │   ├── FRAME_ID_color.png
│   │   ├── FRAME_ID_mask.exr
│   │   ├── FRAME_ID_depth.exr
│   │   ├── FRAME_ID_mask_sam.npz [Optional]
│   │   └── ...
│   └── ...
├── SOPE
│   ├── PATCH_ID
│   │   ├── train
│   │   │   ├── SCENE_NAME
│   │   │   |  ├── SCENE_ID
│   |   |   |  |   ├── FRAME_ID_meta.json
│   │   │   |  |   ├── FRAME_ID_color.png
│   │   │   |  |   ├── FRAME_ID_mask.exr
│   |   |   |  |   ├── FRAME_ID_depth.exr
│   │   │   |  |   ├── FRAME_ID_depth_1.exr [Optional]
│   │   │   |  |   ├── FRAME_ID_coord.png   [Optional]
│   │   │   |  |   ├── FRAME_ID_ir_l.png    [Optional]
│   │   │   |  |   ├── FRAME_ID_ir_r.png    [Optional]
│   │   │   |  |   └── ...
│   │   │   |  └── ...
│   │   │   └── ...
│   │   └── test
│   └── ...
├── PAM
│   └── obj_meshes
│       ├── DATASET-CLASS_ID
│       └── ...
└── Meta    
    ├── obj_meta.json 
    └── real_obj_meta.json

File Description

ROPE

SCENE_ID: The scene ID of the real dataset, e.g., 000000. Each scene consists of multiple frames from a single video.
FRAME_ID_meta.json: The meta-information of the frame.
FRAME_ID_color.png: The color image of the frame.
FRAME_ID_mask.exr: The instance-level segmentation mask of the frame.
FRAME_ID_depth.exr: The depth map of the frame, captured by RealSense D415.
FRAME_ID_mask_sam.npz: The instance-level segmentation mask generated by SAM with a post-processing step. Optional

SOPE

PATCH_ID: The patch ID of the simulated dataset, e.g., 00. We divide the SOPE dataset into 51 patches to facilitate the download and usage of the dataset. Each patch consists of multiple scenes.
train: The training split of the dataset.
test: The testing split of the dataset.
SCENE_NAME: The scene name of the simulated dataset, e.g., ikea. We generate the SOPE dataset with three different scene datasets: ikea, matterport, and scannet++.
SCENE_ID: The scene ID of the simulated dataset, e.g., 0000. Each scene consists of multiple frames generated from different views of the scene.
FRAME_ID_meta.json: The meta-information of the frame.
FRAME_ID_color.png: The color image of the frame.
FRAME_ID_mask.exr: The instance-level segmentation mask of the frame.
FRAME_ID_depth.exr: The depth map of the frame, which is generated with physics-based depth sensor simulation.
FRAME_ID_depth_1.exr: The synthetic perfect depth map of the frame. Optional
FRAME_ID_coord.png: The NOCS map of the frame, proposed by NOCS. Optional
FRAME_ID_ir_l.png: The simulated left infrared image of the frame, generated by the SOPE dataset generation pipeline for depth sensor simulation. Optional
FRAME_ID_ir_r.png: The simulated right infrared image of the frame, generated by the SOPE dataset generation pipeline for depth sensor simulation. Optional

PAM

obj_meshes: The 3D object meshes used in SOPE.
DATASET-CLASS_ID: The class ID of the object in the dataset, e.g., google_scan-book_0168. DATASET is the source of the object, and we provide the pose-aligned 3D models.

Download

We provide two ways to download the dataset:

If you want to preview and download a specific subset of the dataset, you can access it here: Omni6DPose.
If you want to download the entire dataset, you can follow the official instructions at Omni6DPoseAPI.