his service will develop an application that allows citizen science platforms (or other research projects) to develop automatic identification tools adapted to their needs, e.g. for particular groups of species. The identification tool, based on machine learning, will include a search to find similar results. For instance, among photographs of species that potentially resemble the observation made by the citizen scientist.
Example: A museum developing an app related to a particular group of species will be able to integrate a photo-identification tool for that species without much effort.
The application will allow users to define a list of species of interest and, in return, the system will deploy a queryable contextualized identification web service through a secured web API and through a web GUI. Therefore, we will extend and improve the deep-learning-based similarity search engine already in use in the Pl@ntNet platform. This software allows users to identify observations composed of multiple part-based images of the same individual plant (e.g. leaf, flower, fruit, etc.), while returning the most similar observations in the database for each of the identified species, thanks to a scalable similarity-search engine that relies on high-dimensional data hashing and deep representations.
Providing users with accurate visual feedback is crucial so they can control and (un)validate the proposed species.