HyperImageNet: A Large-Scale High-Spatial Resolution Hyperspectral Imagery Classification Benchmark

2026-07-23

AI Summary

Existing hyperspectral image benchmarks lack the capacity to support fine-grained land cover classification and evaluation under open-world conditions. To address this limitation, this work introduces HyperImageNet, a large-scale hyperspectral benchmark comprising 26,084 image patches, 224 spectral bands, and 138 fine-grained land cover categories. Notably, it is the first benchmark to provide both pixel-level semantic labels and instance masks, enabling research in both semantic and instance segmentation. Designed with a spatially disjoint evaluation protocol to simulate open-environment scenarios, HyperImageNet leverages high spatial resolution, rich spectral information, and multi-level annotations to become the largest hyperspectral image classification benchmark to date, significantly advancing fine-grained land cover understanding and open-set remote sensing research.

Abstract

We present HyperImageNet, a large-scale benchmark for fine-grained hyperspectral land-cover understanding. The dataset contains 26,084 airborne hyperspectral image patches with 224 spectral bands and 138 fine-grained land-cover categories. Unlike existing datasets, HyperImageNet provides raw imagery, pixel-level semantic labels, and object-level instance masks, supporting both semantic and instance segmentation. Furthermore, we establish an open-environment benchmark with strict spatial separation to evaluate representative methods and the HyperFree foundation model. Experimental results demonstrate the effectiveness of HyperImageNet for fine-grained hyperspectral understanding and open-environment remote sensing research.

Problem

• hyperspectral imagery
• fine-grained classification
• land-cover understanding
• large-scale benchmark
• open-environment remote sensing

Innovation

• hyperspectral imagery
• fine-grained classification
• semantic segmentation
• instance segmentation
• open-environment benchmark