Claim · model-written summary
GL-MobFormer maintained its performance with a low computational footprint of 0.33 G FLOPs.
Not reportedMaterialGeneCompoundOrgan / stageInteractorInteractor roleSecond interactorSecond interactor roleTreatmentComparatorSettingDirectionQualifier
Verbatim
t0“tea leaf phytopathogen diagnosis”
e0“a low computational footprint of merely 0.33 G FLOPs”
Title
Bridging CNNs and vision transformers for efficient tea leaf phytopathogen diagnosis: GL-MobFormer
Yuntong Yang; Xiaosong Li; Qinzi Li; et al. Frontiers in Plant Science 2026. doi:10.3389/fpls.2026.1814962
Other claims from this paper (3)
- Method / resource GL-MobFormer achieved a classification accuracy of 95.13% on the tea leaf image dataset.
- Method / resource GL-MobFormer reached a Matthews Correlation Coefficient of 0.9417.
- Method / resource Masking only the top 5% of critical activation regions reduced classification confidence by an average of 70.61%, indicating the model relies on lesion features rather than background.