Proposes Two-Weighted Activation Mapping (TWAM) in an Attention-based MIL framework guided by saliency maps that outperforms prior weakly supervised methods by at least 16 points on Citrus Pest and Insect Pest benchmarks while producing location maps without bounding-box labels.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2021 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Weakly Supervised Attention-based Models Using Activation Maps for Citrus Mite and Insect Pest Classification
Proposes Two-Weighted Activation Mapping (TWAM) in an Attention-based MIL framework guided by saliency maps that outperforms prior weakly supervised methods by at least 16 points on Citrus Pest and Insect Pest benchmarks while producing location maps without bounding-box labels.