Dataset · robbyant
Large RGB-D dataset for scene depth estimation (LingBot-Depth)
robbyant/mdm_depth
A massive collection of millions of RGB images and depth maps totaling 2.71 TB for training 3D vision and robotics.
What's inside
The dataset was created for training the LingBot-Depth model and includes several subsets featuring real-world scenes (e.g., 1.4 million samples in RobbyReal). Each example consists of a standard RGB image, raw sensor depth data, and a 16-bit PNG ground truth depth map in millimeters. The total data volume is 2.71 TB.
What you can do with it
- Training and fine-tuning depth estimation and depth map hole-filling models
- Developing computer vision and indoor robot navigation systems
- Testing 3D scene reconstruction algorithms using RGB-D camera data
- Tasks
- Depth estimation
- Languages
- English
- Downloads this month
- 1.9K
- Likes
- 38
How to load it
from datasets import load_dataset
ds = load_dataset("robbyant/mdm_depth")