LVIS-35k: A Large-Scale Visual Instance Recognition Dataset Subset

LVIS-35k is a 35,000-sample subset of a large-scale dataset for instance segmentation and object detection. The dataset was created by Voxel51 and is distributed via the FiftyOne library on Hugging Face. It was last updated on July 8, 2024.

Use Cases

• Training instance segmentation models based on the large-scale visual recognition task.
• Benchmarking object detection algorithms on a dataset with a high number of categories.
• Developing and testing data curation workflows using the integrated FiftyOne tooling mentioned in the description.

Strengths

• Dataset contains 35,000 annotated image samples.
• Provided by Voxel51, a known provider of dataset management tools.

Limitations

• This is explicitly a subset; the full dataset is not included.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the full dataset is unknown, which may limit suitability assessment.

Provenance

Source

Voxel51 via Hugging Face.

Collection Method

Subset created from the full LVIS dataset; a notebook recipe for creation is referenced.

Freshness

Last updated 2024-07-08 19:49:51; freshness should be verified.