TAO-Amodal: Amodal Bounding Box Annotations for Occluded and Out-of-Frame Objects

TAO-Amodal augments the TAO dataset with amodal bounding box annotations for fully invisible, out-of-frame, and occluded objects. The dataset also includes modal segmentation masks. It was created by Cheng-Yen (Wesley) Hsieh and was last updated on the Hugging Face platform on January 11, 2025.

Use Cases

• Train models for amodal object detection based on annotations for occluded and out-of-frame objects.
• Benchmark scene understanding algorithms using the provided modal segmentation masks.
• Research object permanence and tracking in video sequences using the amodal bounding box labels.

Strengths

• Extends the established TAO dataset with specialized amodal annotations.
• Includes both amodal bounding boxes and modal segmentation masks, as described.

Limitations

• Dataset size, row count, and file formats are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last updated 2025-01-11 19:02:55; freshness should be verified.

Provenance

Source

chengyenhsieh on Hugging Face, augmenting the TAO dataset.

Collection Method

Augmentation of an existing dataset with manual or semi-automated amodal annotations.

Freshness

Last updated 2025-01-11 19:02:55.