MAZE Image Processing Pipeline

The MAZE Image Processing Pipeline (MAZE-IPP) is a comprehensive tool for advanced marine biological research, integrating three specialized modules to enhance zooplankton imagery analysis. Its command-line interface (CLI) provides two modules: The first module (maze-ipp loki) resegments LOKI (Lightframe On-sight Keyspecies Investigation) data for precise organism delineation. The second module (maze-ipp predict) applys a deep learning model to images for semantic segmentation or classification.

MAZE-IPP interfaces with MorphoCluster and EcoTaxa, two image annotation platforms. MorphoCluster’s clustering algorithms accelerate the annotation of images using the deep image features calculated by maze-ipp predict. Segmented and classified images are exported to EcoTaxa for further validation and curation. Fine-grained polyhierarchical identification of calanoid copepods and chaetognatha, detailing orientation, health, life stage, and sex is provided using the polytaxo library. Interfacing with EcoTaxa is implemented using pyecotaxa.

Interaction of the pipeline components within the EcoTaxa ecosystem.