Bi-temporal Satellite Imagery With Change Captions

Levir Mci contains bi-temporal satellite images paired with change detection masks and descriptive sentences. It was created by author lcybuaa and last updated on HuggingFace in November 2024. The dataset supports multi-task learning for visual change analysis.

Use Cases

• Train a model to generate descriptive sentences from pairs of bi-temporal images.
• Develop a segmentation model to predict change detection masks from image pairs.
• Benchmark multi-task learning architectures that perform simultaneous change detection and captioning.
• Fine-tune vision-language models on the specific domain of geospatial change analysis.

Strengths

• Includes diverse change detection masks for segmentation tasks.
• Provides descriptive sentences enabling vision-language applications.
• Supports multi-task learning for change detection and captioning.

Limitations

• Specific row count, column details, and dataset size are unknown.
• Geographic coverage and temporal range of the imagery are unspecified.
• The license terms for use and redistribution are not provided.

Provenance

Source

lcybuaa on HuggingFace, associated with the Change-Agent GitHub repository.

Collection Method

Collection of bi-temporal satellite images annotated with masks and captions.

Freshness

Last updated on HuggingFace in November 2024.