MSCOCO: Large-Scale Object Detection, Segmentation, and Captioning Dataset

The MSCOCO dataset contains over 330,000 images, with more than 200,000 labeled. It includes 1.5 million object instances across 80 object categories and 91 stuff categories for segmentation. Each image has 5 captions, and the dataset contains annotations for 250,000 people with keypoints.

Use Cases

• Train object detection models based on 1.5 million annotated object instances across 80 categories.
• Develop image segmentation models based on superpixel stuff segmentation for 91 stuff categories.
• Build image captioning systems based on 5 captions provided for each image.
• Train human pose estimation models based on keypoint annotations for 250,000 people.

Strengths

• Large scale with over 330,000 images.
• Multi-task annotations supporting object detection, segmentation, and captioning.
• Includes 1.5 million object instances and keypoints for 250,000 people.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface user shunk031

Freshness

Last updated 2024-06-14 07:16:29; freshness should be verified.