An OpenAI model for mass processing of trap liners as part of an area-wide IPM program

Various companies are developing and offering camera traps and other Internet of Things (IoT) devices for monitoring insect pests. One barrier to use for these devices in the western US is poor cell signals in the rural areas they are needed. Expense to replace cheap many traditional traps is another concern. The current project uses annotation of images from traditional sticky traps to tune a general-purpose model (YOLO) for the navel orangeworm, Amyelois transitella (Walker)(Lepidoptera: Pyralidae) (NOW), which is a key pest of almonds, pistachios, and walnuts. The resulting tuned model can be used with edge computing devices (a camera-attached PC or single-board computer) to automate identification and counting of NOW on many liners generated at regular intervals in an area-wide program. In addition, the image library and the tuned model will be shared publicly as a common resource for the tree nut industry and its service providers.