VisualData Discovery is one of the largest public repositories of compute vision datasets. It currently indexes over 500 high quality datasets created by the vision community including universities, companies and hobbyists. People around the world visit VisualData everyday to find the best datasets for their project.
From day 1, VisualData Discovery was built to be engaging and useful for the community. To manage the quality and correctness of the information for each dataset, the sourcing and ingestion of datasets have been made exclusively internally by a semi-automated process. In the past year or so, we received a lot of requests to add specific datasets to the site. It only makes sense to give everyone the ability to share datasets with the public in a central place. We are excited to open up the Dataset Creation Editor where you can contribute a dataset to VisualData Discovery with any information you would like to use. Let’s see how to do that.
Before adding a new dataset, use the search bar and topic selector to check if the dataset already exists. An error will occur if you try to add a duplicated dataset. To add the dataset, click ‘ADD DATASET’ button in the middle of the toolbar. You will see an editor to input dataset information like one shown below (login is required).
The editor is self-explanatory with tips under each field. The one with ‘*’ is a required field and can not be skipped. After filling in all necessary information, click ‘Confirm’ at the bottom to submit the dataset. The submission will go through a review process and be approved (normally within 24 hours) if no issue is found. The contributor’s name will be shown in the dataset detail as a token of appreciation. If you have any question or feedback, you can use the blue button on the bottom right to send us a message.
We are dedicated to make it easier to discover and share datasets for the computer vision community and facilitate research and development of new techniques to solve problems and build applications. Can’t wait to see what datasets you will share with the world!