31,500 images across 45 distinct scene classes, each containing 700 examples, form this benchmark for remote sensing. Created by Northwestern Polytechnical University, the dataset is hosted on Hugging Face and was last updated in January 2024. It is designed for the task of Remote Sensing Image Scene Classification (RESISC).
Use Cases
- Benchmarking scene classification models based on the 45 distinct land-use and land-cover categories
- Training remote sensing image classifiers based on the 700 images per class
- Developing and evaluating transfer learning techniques for geospatial imagery
- Studying model performance on a standardized, publicly available remote sensing dataset
Strengths
- Contains 31,500 total images, providing a substantial volume for model training
- Offers a balanced set with 700 images per each of the 45 scene classes
- Serves as a publicly available benchmark created by a recognized academic institution
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download
- Column-level documentation is absent; field semantics must be inferred after download
Provenance
- Source
- Northwestern Polytechnical University (NWPU)
- Freshness
- Last updated 2024-01-07 18:11:08; freshness should be verified