Skip to content

Latest commit

 

History

History
56 lines (39 loc) · 2.81 KB

README_dataset.md

File metadata and controls

56 lines (39 loc) · 2.81 KB
Raw

Egocentric Depth on everyday INdoor Activities (EDINA) Dataset

🌟 Full EDINA dataset is now available to download! 🌟

Overview

EDINA is an egocentric dataset that comprises more than 500K synchronized RGBD frames and gravity directions. Each instance in the dataset is a triplet: RGB image, depths and surface normals, and 3D gravity direction.

edina2.gif

Capturing Process

The data were collected using Azure Kinect cameras that provide RGBD images with inertial signals (rotational velocity and linear acceleration). Eighteen participants were asked to perform diverse daily indoor activities, e.g., cleaning, sorting, cooking, eating, doing laundry, training/playing with pet, walking, shopping, vacuuming, making bed, exercising, throwing trash, watering plants, sweeping, wiping, while wearing a head-mounted camera. More information can be found in our main paper and supplementary materials.

Dataset Downloading

Raw data

We provide a Python script download_edina.py to download the dataset conveniently. More information on dataset format can be found in the Data Organization section.

To download the dataset to pathToDataset AND unzip, you can use the following command by specifying the --split argument to be either train or test to download the corresponding train/test data:

python3 download_edina.py --out_dir pathToDataset --split test --unzip

We also provide optional functionalities where you can only download a specific scene (e.g., scene0016_01):

python3 download_edina.py --out_dir pathToDataset --id scene0016_01 --unzip

Please refer to download_edina.py for more specific details.

Dataset splits

We provide the .pkl file that specifies the train/test split of our data, specified by the dictionary keys edina_train and edina_test. The pickle file can be downloaded to ./pickles/ folder by:

wget -O scannet_edina_camready_final_clean.pkl https://edina.s3.amazonaws.com/pickles/scannet_edina_camready_final_clean.pkl && mv scannet_edina_camready_final_clean.pkl ./pickles/

Data Organization

There is a separate directory for each RGB-D-(Normal-Depth) sequence (with varied length). Each sequence is named uniquely in the format of scene<participantID>_<videoID>, or scene%04d_%02d, sorted sequentially by participantID (from 0 to 17) and videoID (0-indexed) per participant. Within each sequence, <frameID> is also 0-indexed and of format %06d.

The general data hierarchy is described below:

scene<participantID>_<videoID>
├── color
│   └── color_<frameID>.png
├── depth
│   └── depth_<frameID>.png
├── normal
│   └── normal_<frameID>.png
├── gravity
│   └── gravity_<frameID>.txt