Lightweight YOLOv8s-BD for small and dense agricultural pest detection using dynamic upsampling and bidirectional feature fusion

Qiang Gao, Chongchong Shi, Yu Ji, Xiaoning Chen, Huan Wang, Mengke Cao, Meili Wang · 2026

Read Lightweight YOLOv8s-BD for small and dense agricultural pest detection using dynamic upsampling and bidirectional feature fusion by Qiang Gao, Chongchong Shi, Yu Ji, Xiaoning Chen, Huan Wang, Mengke Cao, Meili Wang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13616948.txt.

Description

Qiang Gao and six colleagues build a lightweight computer-vision model that spots crop pests in field photos, where insects are tiny, crowded, and easy to confuse with each other. Their model, YOLOv8s-BD, adapts a standard real-time detector for farm use. Rare pest classes were boosted with generated images so the model would not ignore them.

Two design changes do most of the work: a dynamic upsampling step that keeps fine detail for small pests, and a weighted two-way feature pyramid that blends detail across scales. On the Pest24 dataset the model reached 72.3 percent mean average precision on the looser test while staying small and fast, at 7.4 million parameters and under two milliseconds per image. Error checks showed fewer mix-ups between similar species, which matters for choosing the right treatment in the field.

Catalog record

CREATORQiang Gao, Chongchong Shi, Yu Ji, Xiaoning Chen, Huan Wang, Mengke Cao, Meili Wang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE46 KB
LICENSEhttps://creativecommons.org/licenses/by/4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/article/MED/PMC13616948

Browse the full library · About the Tower of Babel