USE CASE 2 / FIVE DATASETS · 50 PHOTOGRAPHS
Five settings.
One closer look.
Application test across vegetation, fungal hyphae, aerial ecosystems, coral reefs, and aquatic microscopy. Explore each dataset independently, with ten photographs and the same local-focus workflow.
01 / 05 · TEN EXAMPLES
Vegetation
Overlapping leaves, stems, and fine textures in dense field vegetation.
MOIN grassland photographs ↗Clustering uses all 16 retained dimensions, independently of the displayed colors. K-means selects its count automatically; GMM uses full covariance with λ=4. How the count is chosen ↗
02 / 05 · TEN EXAMPLES
Fungal hyphae
Fine branching networks, filament edges, and the texture between them.
Fungal Network Analysis ↗Clustering uses all 16 retained dimensions, independently of the displayed colors. K-means selects its count automatically; GMM uses full covariance with λ=4. How the count is chosen ↗
03 / 05 · TEN EXAMPLES
Aerial ecosystems
Crowded crowns and canopy texture. These examples are airborne imagery, rather than UAV acquisitions.
NEON tree benchmark ↗Clustering uses all 16 retained dimensions, independently of the displayed colors. K-means selects its count automatically; GMM uses full covariance with λ=4. How the count is chosen ↗
04 / 05 · TEN EXAMPLES
Coralscapes
Interwoven corals, algae, substrate, and underwater community structure.
Coralscapes reef communities ↗Clustering uses all 16 retained dimensions, independently of the displayed colors. K-means selects its count automatically; GMM uses full covariance with λ=4. How the count is chosen ↗
05 / 05 · TEN EXAMPLES
Aquatic microscopy
Microscopic organism shapes, internal texture, and background contrast.
PMID2019 phytoplankton ↗Clustering uses all 16 retained dimensions, independently of the displayed colors. K-means selects its count automatically; GMM uses full covariance with λ=4. How the count is chosen ↗
From visual detail
to a testable method.
These are feature groups, rather than identified species or verified organism instances. The fifty-image collection is illustrative. Inspect the underlying regularization sweep, count distributions, and repeatability before interpreting a partition biologically.