Human Body 3D Instance Segmentation With 22 Landmarks

Aggregating 18,880 images of 466 people, annotated with 3D instance segmentation masks and 22 anatomical landmarks. It was created by Nexdata and includes diversity in scenes, lighting, ages, shooting angles, and poses.

Use Cases

• Train a 3D instance segmentation model on human body images to separate individuals from complex backgrounds.
• Develop a pose estimation system using the 22 anatomical landmarks annotated on each human body.
• Analyze the impact of varying light conditions and shooting angles on landmark detection accuracy.
• Fine-tune a model for human body parsing using the provided instance segmentation annotations across diverse poses.

Strengths

• Contains 18,880 annotated images, providing a substantial sample for model training.
• Features 466 distinct individuals, contributing to subject diversity.
• Includes annotations for both 3D instance segmentation and 22 landmarks, offering multi-task learning potential.
• Captures diversity in multiple factors: scenes, light conditions, ages, shooting angles, and poses.

Limitations

• The dataset is described as a sample of a larger paid dataset, which may limit its completeness.
• The specific distribution of images per person is unknown, which could affect representation.
• No information is provided on the resolution, format, or consistency of the 18,880 images.

Provenance

Source

Nexdata

Collection Method

null

Time Range

null

Freshness

The dataset was last updated on 2025-04-24.

Geography

null