Abstract
to-do: add abstract
Dataset Overview

HyperMed-X contains 10,000 hyperspectral cubes collected from 15 different tissue types, with full expert annotations for classification, detection, and segmentation tasks. The data was collected using a custom-built hyperspectral imaging system with the following specifications:
Property | Specification |
---|---|
Spatial resolution | 1024 × 1024 pixels |
Spectral range | 450-950nm |
Spectral resolution | 3.9nm |
Number of bands | 128 |
Bit depth | 12-bit |
Data Collection
The data collection process followed a rigorous protocol approved by the institutional review boards of participating hospitals. Patient consent was obtained for all samples included in the dataset, and all data has been anonymized to protect patient privacy.
Imaging was performed ex-vivo immediately after surgical extraction to preserve tissue properties. Calibration procedures were implemented to ensure spectral and radiometric accuracy across all samples.
Annotations
Benchmark Results
We established baseline performance metrics using several state-of-the-art deep learning architectures. The following table summarizes the results on the main tasks:
Model | Classification Accuracy | Detection mAP | Segmentation mIoU |
---|---|---|---|
ResNet-18 | 87.3% | 79.8% | 72.5% |
ResNet-34 | 85.1% | 77.6% | 78.9% |
ResNet-50 | 89.4% | 82.1% | 75.6% |
This dataset is provided for academic and research purposes only. Any commercial use requires explicit written permission from the authors. Researchers are required to cite the corresponding paper in any resulting publications.
Citation
Acknowledgements
This work is supported by funds from the German Federal Ministry of Food and Agriculture (BMEL), based on a decision of the Parliament of the Federal Republic of Germany. The German Federal Office for Agriculture and Food (BLE) provides coordinating support for artificial intelligence (AI) in agriculture as the funding organisation, grant number 28DK116C20. The dataset was created as part of the research project “KIRa - KI-gestützte Plattform zur Klassifikation und Sortierung von Pflanzensamen: Bewertung der Saatgutreinheit am Musterfall Raps" (engl. "AI-supported platform for classifying and sorting plant seeds: Evaluation of seed purity using oilseed rape as a model case").