Floralens: a Deep Learning Model for the Portuguese Native Flora

arXiv:2403.12072v1 Announce Type: cross
Abstract: Machine-learning techniques, namely deep convolutional neural networks, are pivotal for image-based identification of biological species in many Citizen Science platforms. However, the construction of critically sized and sampled datasets to train the networks and the choice of the network architectures itself remains little documented and, therefore, does not lend itself to be easily replicated. In this paper, we develop a streamlined methodology for building datasets for biological taxa from publicly available research-grade datasets and for …

abstract architectures arxiv citizen science construction convolutional neural networks cs.cv cs.lg datasets deep learning flora however identification image machine network networks neural networks pivotal platforms science species train type