Lightweight CNN SE transformer for robust weed classification with optimizer aware performance

Priyadharshini G, Vetriselvi T · 2026

Read Lightweight CNN SE transformer for robust weed classification with optimizer aware performance by Priyadharshini G, Vetriselvi T (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/PMC13493912.txt.

Description

Priyadharshini G and Vetriselvi T build in 2026 a featherweight vision model that spots weeds among young crops for farm robots. Their architecture braids a compact convolutional network with attention blocks that reweight important features, then a transformer stage that reads the scene as a sequence. Depth, skip connections, and training settings were each varied to find the cheapest design that still sees clearly.

Two weed datasets, one with three classes and one with sixteen, put the design through its paces. The Adam training optimizer delivered the best test accuracy at ninety-seven percent, slightly ahead of its label-smoothed variant and clearly ahead of RMSProp. Heat maps confirmed the model stares at the weed itself rather than the soil around it. Precise early weeding could cut herbicide use while sparing crops.

Catalog record

CREATORPriyadharshini G, Vetriselvi T
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE76286
LICENSEhttps://creativecommons.org/licenses/by/4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/article/MED/PMC13493912

Browse the full library · About the Tower of Babel