A new preprint on arXiv addresses the problem of detecting visible surface damage on banana fruits caused by Basilepta subcostata, the banana fruit scarring beetle. Such damage can lower the market value of the fruit, but the current practice of manual inspection is subjective and does not scale well.

The study presents a deep learning-based approach for detecting this damage, according to the abstract. However, the abstract text cuts off mid-sentence, so the specific architecture, dataset, and reported accuracy are not available in the source.

Because the source is incomplete, this digest can only confirm the problem statement and the general direction of the proposed solution. Readers interested in implementation details will need to consult the full paper.