Potential research project directions
3D floral reconstruction and phenotyping
How can photographs become three-dimensional phenotypes for studying floral evolution?
Flowers combine curved surfaces, thin structures, and intricate spatial arrangements. Reconstructing this diversity in three dimensions opens new ways to study floral evolution and function, while posing challenges for computer vision and biological measurement.
Our turntable-based workflow supports automated batch photogrammetric reconstruction of flowers from NYBG’s Living Collections. With these images and 3D reconstructions, we explore photogrammetry, neural reconstruction, and generative 3D. We also investigate landmark-free approaches and machine learning to represent floral form and compare shapes across specimens and species.
- Reconstructing complex flowers. Recovering intricate floral geometry with image-based and generative methods while remaining faithful to the specimen.
- Learning representations of floral form. Making shapes comparable across specimens and species through landmark-free descriptors and learned representations.
- Connecting methods to evolution. Using phylogenetic comparisons to study floral diversification, trait integration, and the repeated evolution of similar forms.