Original title: Using Deep Learning for Image-Based Plant Disease Detection
The researchers train a computer program to recognize plant leaves in photos. It learns from 54,306 images showing healthy leaves and leaves with diseases, collected under controlled conditionsImages collected in a managed setting rather than under the changing conditions of everyday farms.. The program identifies 14 crop species and 26 diseases, or whether a disease is absent. On a separate test set, it is correct 99.35% of the time, suggesting this approach could support diagnosis with smartphones.
Imagine sorting a pile of leaves into groups by crop and by whether they look healthy. The tricky part is learning which markings matter when many leaves can look similar. In this paper, the pile is made of leaf photos, and the sorting practice teaches a program to identify crop types and diseases from new photos.
Farmers in places without easy access to disease-diagnosis infrastructure may be able to use smartphone photos to help identify crop diseases. Earlier identification could support food security, though the abstract tests images collected under controlled conditions, not everyday farm use.
The abstract reports results on images collected under controlled conditions and a held-out test set. It does not establish performance on photos taken in everyday farm settings, or show that smartphone diagnosis improves crop outcomes. The reported accuracy applies to the test set described.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Using Deep Learning for Image-Based Plant Disease Detection — Sharada P. Mohanty, David Peter Hughes, Marcel Salathé — Frontiers in Plant Science — doi:10.3389/fpls.2016.01419
Abstract licence: CC BY
Explanation generated by Synthesis from the abstract — it is not the original paper.