RESISC45 Overview Usage Dataset Information The dataset is divided into the following splits: Training set : Contains 18,900 examples, used for model training. Test set : Contains 6,300 examples, used for model evaluation and benchmarking. The dataset also includes the following augmented sets, which can be used for testing the model's robustness to various types of image corruptions: Contrast enhanced set : Contains 6,300 examples with enhanced contrast for improved feature visibility. Gaussian noise set : Contains 6,300 examples where images have been corrupted with Gaussian noise. Impulse noise set : Contains 6,300 examples with impulse noise. JPEG compression set : Contains 6,300 examples where images have been compressed using JPEG encoding. Motion blur set : Contains 6,300 examples with motion blur applied. Pixelate set : Contains 6,300 examples where images have been pixelated. Spatter set : Contains 6,300 examples with spatter noise.
Runs entirely in your browser via DuckDB-Wasm — this dataset's real data file is loaded once, then queried locally. Nothing is sent to a server.
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy