?url_ver=Z39.88-2004&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Adc&rft.title=Dual-attention+Focused+Module+for+Weakly+Supervised+Object+Localization&rft.creator=Zhou%2C+Y&rft.creator=Chen%2C+Z&rft.creator=Shen%2C+H&rft.creator=Liu%2C+Q&rft.creator=Zhao%2C+R&rft.creator=Liang%2C+Y&rft.description=The+research+on+recognizing+the+most+discriminative+regions+provides+referential+information+for+weakly+supervised+object+localization+with+only+image-level+annotations.+However%2C+the+most+discriminative+regions+usually+conceal+the+other+parts+of+the+object%2C+thereby+impeding+entire+object+recognition+and+localization.+To+tackle+this+problem%2C+the+Dual-attention+Focused+Module+(DFM)+is+proposed+to+enhance+object+localization+performance.+Specifically%2C+we+present+a+dual+attention+module+for+information+fusion%2C+consisting+of+a+position+branch+and+a+channel+one.+In+each+branch%2C+the+input+feature+map+is+deduced+into+an+enhancement+map+and+a+mask+map%2C+thereby+highlighting+the+most+discriminative+parts+or+hiding+them.+For+the+position+mask+map%2C+we+introduce+a+focused+matrix+to+enhance+it%2C+which+utilizes+the+principle+that+the+pixels+of+an+object+are+continuous.+Between+these+two+branches%2C+the+enhancement+map+is+integrated+with+the+mask+map%2C+aiming+at+partially+compensating+the+lost+information+and+diversifies+the+features.+With+the+dual-attention+module+and+focused+matrix%2C+the+entire+object+region+could+be+precisely+recognized+with+implicit+information.+We+demonstrate+outperforming+results+of+DFM+in+experiments.+In+particular%2C+DFM+achieves+state-of-the-art+performance+in+localization+accuracy+in+ILSVRC+2016+and+CUB-200-2011.&rft.publisher=ArXiv&rft.date=2019-09-11&rft.type=Working+%2F+discussion+paper&rft.language=eng&rft.source=++++ArXiv+(2019)+++++&rft.format=text&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F10137826%2F1%2F1909.04813v1.pdf&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F10137826%2F&rft.rights=open