Dataset · mickeykang
Real-world recordings of 16 household objects for 6D pose estimation (Event6D)
mickeykang/Event6D
A test dataset from a CVPR 2026 paper featuring handheld recordings of everyday objects for detection tasks.
What's inside
The repository contains fewer than a thousand records with images and text metadata captured by hand in real conditions for the Event6D (CVPR 2026) project. The dataset includes recordings of 16 everyday objects in the style of the YCB dataset from random viewpoints and with arbitrary motion. The data is split into a training set (37 sequences, 16 objects) and a primary test split (14 sequences, 14 objects).
What you can do with it
- Evaluating detection and 6D object pose estimation models trained on synthetic data (sim-to-real)
- Testing object recognition algorithms under arbitrary camera motion
- Tasks
- Object detection
- Size
- up to 1K rows
- Downloads this month
- 57.1K
- Likes
- 1
How to load it
from datasets import load_dataset
ds = load_dataset("mickeykang/Event6D")