Omnidata Live Demo

We provide a demo where you can upload (or use an existing) query and see the results of different Omnidata models vs various baselines.

The Omnidata models were trained using cross-task consistency and 3D data augmentations. You can download the pre-trained models here.

You can also visit the archive page to examine previously uploaded query images by online users and their corresponding results. The images you upload will be added to this archive too, unless you ask us to remove them.

The networks were trained on several million images of the starter dataset, featuring mostly general indoor scenes (rather than faces, humans, landscapes, etc). If your query image severely deviates from the training data, the performance is expected to degrade.

Running the full demo usually takes around 20 seconds, depending on the traffic.


Omnidata models last updated: 24 March 2022

Upload your own image or click on one of the sample queries below. Click on the cube to use a random query image from previous uploads.

Sample 1 Sample 2 Sample 3 random
If you find the pre-trained models useful, please cite our papers: 
@inproceedings{eftekhar2021omnidata,
	title={Omnidata: A Scalable Pipeline for Making Multi-Task Mid-Level Vision Datasets From 3D Scans},
	author={Eftekhar, Ainaz and Sax, Alexander and Malik, Jitendra and Zamir, Amir},
	booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
	pages={10786--10796},
	year={2021}
}
@inproceedings{kar20223d,
	title={3D Common Corruptions and Data Augmentation},
	author={Kar, O{\u{g}}uzhan Fatih and Yeo, Teresa and Atanov, Andrei and Zamir, Amir},
	booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
	pages={18963--18974},
	year={2022}
}