Investigating State of the Art Hyperspectral Imaging Classification Models for Plastic Types Identification

Polymers Dataset

Description:

The polymers dataset is a multiscene Hyperspectral benchmark dataset comprising of reference polymer samples and shredded polymer samples in the visible to short-wave infrared, capturing 450 bands within the [400–2500] nm range using an AisaFENIX (Spectral Imaging Ltd, Oulu, Finland) spectrometer. 

Two sample batches were investigated:

reference polymers of known composition and dimensions (15 X 10) cm, commonly found in e-waste. shredded pieces of polymers with sizes ranging from (0.3–4) cm.

Data Format

HSI data: each hyperspectral data cube is accompanied by a data file and a .hdr file. Ground truth mask: .png file (only for multi samples scenes)

Folder Organization

Polymers

Test

HSI: .dat & .hdr Ground truth mask: test.png False colour representation of the scene: .png

Train

HSI_ : 3 different scans of reference samples scanned PC, PE, PET, PP: Hyperspectral cubes (11x11x450) .hdr & .dat

Data Classes in Masks

Masks contain 1 to 6 segmentation classes:

1: "PP" 2: "Black Plastic" 3: "PVC" 4: "PET" 5: "ABS" 6: "PE"

Code Repository

To facilitate reading and working with the data, Python codes are available on the GitHub repository:

https://github.com/hifexplo

https://github.com/Elias-Arbash

Citation

If you use this dataset, please cite the following article: (To be filled once published)

Contact

For further information or inquiries, please visit our website:

https://www.iexplo.space/

Contact Email: e.arbash@hzdr.de

The HSI dataset of the work: INVESTIGATING STATE OF THE ART HYPERSPECTRAL IMAGING CLASSIFICATION MODELS FOR PLASTIC TYPES IDENTIFICATION