Solve the taxonomic impediment of plants in Lebanon by developing an accessible and real-time plant recognition tool, in the form of a smartphone application for stakeholders, based on computer vision and ML.
Methods and Technologies
Collecting data on Lebanese plant species from different sources, i.e., data mining including manual and automated data entry from online and offline databases, finding gaps in existing database sources such as species locations, features (such as species flower color, flowering time, canopy shape…), species gena and families, climate and weather data.
Developing and/or training machine learning models on a combined Lebanese dataset: development and testing of machine learning models for plant species identification, e.g., neural networks, support vector machines, decision trees; training and testing on open-access international datasets as well as Lebanon-specific data, including the application of transfer learning.
Groundwork for the application creation, e.g., input specification, algorithm adoption, output format, and storage.
Conducting a pilot study in the Shouf Biosphere Reserve (SBR), documentation (including labeling) of the data collected during field trips.
Conducting stakeholder interviews following the principles of human-centered design.
