Microscopy images of potassium permanganate–treated edible oils (seven classes, 111 preparations) for reaction-assisted microscopy and deep learning–based oil identification
Year: 2026
Type: other
Abstract: This dataset contains light-microscopy images of edible oils and oil blends after oxidation with potassium permanganate (KMnO4), together with the partition manifests and analysis code used to evaluate image-based oil identification at the level of independent sample preparations. Classes (folder → oil): NutOil → hazelnut oil; Olive Oil → olive oil; Riviera → Riviera olive oil; Sunflower → sunflower oil; Olive_Nut, Olive_Sunflower, Sunflower_nut → 1:1 (v/v) blends. For each preparation, 30 mL of oil were mixed with 5 mL of 0.1 N aqueous KMnO4 and vortexed for 1 min; a drop was imaged immediately without a coverslip with an Inskam 307 digital microscope (nominal 1000×, automatic white balance). Three fields were photographed per preparation: 345 RGB JPEG images (640 × 480) from 111 preparations. frames_manifest.csv and patches_manifest.csv give the preparation ID and the test/cross-validation fold of every frame and tile; the notebook reproduces all analyses. Important for re-use: frames and tiles from the same preparation are not independent; splits should be made at the preparation level (prep_id). Limitations: one retail bottle per oil type, one microscope, one blend ratio.
Author: Yılmaz, Burak
Institution: Konya Technical University
Language: English
Subfield: Analytical Chemistry
Field: Chemistry
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7220909679