Lecture: 'Image Classification in Low Data Environments'

Elmer Fernández (CONICET - Catholic University of Córdoba, Argentina)

Image diagnosis has generally relied on complex algorithms for their manipulation, with varying levels of success. In this context, the versatility of neural network algorithms based on the deep learning paradigm is an excellent alternative, given the versatility they present. However, the optimization process for these networks requires a quantity of data that significantly exceeds what can be obtained in many applications and environments, leaving this possibility available either for large companies or for large specialized centers. One alternative for this type of problem is the possibility of using neural networks trained in other contexts as general feature extractors from images and using these features as input to another classification model whose optimization does not require a large number of examples.

More info: https://citius.gal/en/events/conference-image-classification-in-low-data-quantity-environments/