Tutorial / Discover & Reuse

How to Find the Right Dataset

Define Dataset Search Criteria

Before searching, define what matters most for your use case. Consider the required modality, task, license, release year, and object classes. In this example, we are looking for a recent object-detection dataset with a clearly specified license.

Criteria for choosing a dataset, including modality, task, license, release year, and object classes

Enter Keywords in the Homepage Search

From the OpenConstruction homepage, enter a task or topic such as object detection. The search returns candidate resources related to those keywords.

OpenConstruction homepage search field highlighted

Open the Dataset Catalog

You can also go directly to the Dataset Catalog. Open the Catalog menu and select Datasets.

Catalog navigation menu with Datasets highlighted

Use the Dataset Count Shortcut

As another entry point, select the homepage indicator showing the number of available datasets to open the full dataset catalog.

Dataset count shortcut highlighted on the OpenConstruction homepage

Start with a Simple Search

The Dataset Catalog provides search, filter, and sorting controls. Begin with a simple task description, such as object detection.

Dataset Catalog with search, filter, and sorting controls

Search by Name, Class, or Term

Search using a dataset name, task, object class, or another related term. Use the language most closely associated with your intended application.

Search field in the Dataset Catalog

Run the Search

After entering your terms, select the search action to query the OpenConstruction catalog.

Search action highlighted in the Dataset Catalog

Search for Object Detection

For this example, enter object detection and press Enter or use the search action.

Object detection entered in the Dataset Catalog search field

Review Candidate Datasets

The catalog returns candidate datasets matching the search. Review the initial results, then use filters to narrow the list.

Dataset Catalog results for an object detection search

Use the Filter Panel

Refine the results by modality, task, license, release year, and object class. Combine only the criteria necessary for your use case.

Dataset Catalog filter panel highlighted

Filter by Modality

Select the data type you need, such as 3D point clouds or aerial RGB imagery, to exclude incompatible datasets.

Modality options in the Dataset Catalog filter panel

Filter by Task and License

Next, narrow the results by task and license. Review the license terms carefully to confirm that the permitted use aligns with your project.

Task and license options in the Dataset Catalog filter panel

Assess the Filter Impact

Watch the result count as you apply filters. The changing count helps indicate whether the search is still too broad or has become too restrictive.

Filtered dataset result count highlighted

Sort the Remaining Results

Use Sort by to organize the remaining datasets by year, added date, name, sample count, or number of classes. To find a recent dataset, sort by year in descending order.

Sort control highlighted above filtered dataset results

Open the Dataset Details

Select View details to inspect the complete record for a promising dataset.

View details action highlighted on a dataset result

Verify the Dataset Information

Before deciding to use the dataset, confirm that its task, modality, license, release information, and original source align with your requirements. Follow the provider link to review the authoritative access and license terms.

Dataset detail page showing imagery, metadata, license, and source access

Return to the Catalog

If the dataset is not a good fit, return to the catalog and adjust the search terms or filters. Finding the right dataset is usually an iterative process.

Back to catalog action highlighted on a dataset detail page

Refine the Search Iteratively

Search, filter, verify, and refine until you identify a strong candidate. This process moves you from a broad idea to a relevant dataset with clear supporting information.

Tutorial conclusion stating that a strong dataset candidate has been found

You are now ready to search the OpenConstruction dataset catalog with clear criteria and evaluate each candidate carefully.