A preprint posted on arXiv describes a planned comparison of five convolutional neural network (CNN) architectures for diagnosing branch dieback disease in pequi trees. The disease, known locally as seca-dos-ponteiros, affects Caryocar species. The study specifically lists AlexNet, Inception, LeNet, VGG16, and MobileNet as the models to be evaluated.

The abstract, however, only states the objective of the research. It does not include any results, accuracy figures, or conclusions about which architecture performs best. Because this is the sole source, there are no conflicting findings to reconcile.

For now, the work signals an application of deep learning to a specific agricultural problem: detecting a branch disease in a tropical tree. Full results will presumably appear in a later version of the paper or in a peer-reviewed venue.