Fig. 4: VespAI hardware and performance.
From: VespAI: a deep learning-based system for the detection of invasive hornets
a Diagram of components and optional additions for the detector hardware. The system is built around a Raspberry Pi 4, with flexible modular components including (1) a 16MP IMX519 autofocus camera module; (2) a 4 G HAT with GNSS positioning for remote transmission of detections via SMS; (3) a PiJuice 12,000 mAh Battery; and (4) a PiJuice 40 watt solar panel for self-sustaining remote deployment. The hardware configuration is not limited to these components and will work with any Raspberry Pi 4-compatible additions, allowing for complete customisation based on use case and budget. Photographs courtesy of Raspberry Pi Ltd. b Prototype setup of bait station and hardware to test the VespAI algorithm in the field. c, d Precision and recall scores across candidate cameras for c V. velutina and d V. crabro during field testing of the prototype system (N = 55). Boxplots are coloured by measure (precision, blue; recall, grey) and grouped by camera performance. Dashed lines indicate the desirable precision threshold of >0.99. Outliers (greater than 1.5 times the interquartile range from the median) are denoted with circles.
