Feature pyramid networks for object detection

Transcript

1. ͲΜͳ΋ͷʁ ઌߦݚڀͱൺ΂ͯԿ͕͍͢͝ʁ ٕज़ͷख๏΍؊͸ʁ վળͷ༨஍͸͋Δʁ Ͳ͏΍ͬͯ༗ޮͩͱݕূͨ͠ʁ ࣍ʹಡΉ΂͖࿦จ͸ʁ ɾW. Liu, D. Anguelov, D. Erhan, C. Szegedy, and S. Reed. SSD: Single shot multibox detector. In ECCV, 2016. ಛʹͳ͠ ɾ80ΧςΰϦͷCOCO de-tectionσʔληοτΛ༻͍࣮ͯݧ ɾ໿2ϙΠϯτͰ͸͋Δ͕ɺैདྷख๏ͱൺֱͯ͠վળ͞Εͨ ɾ೚ҙͷαΠζͷγϯάϧεέʔϧը૾Λೖྗͱ͠ɼͦΕʹൺྫͨ͠αΠζͷಛ௃Ϛο ϓΛग़ྗ͢Δ ɾConvNetΛը૾ϐϥϛουͷεϥΠσΟϯά΢Οϯυ΢ݕग़ثͱͯ͠ద༻͢Δ͜ͱ Ͱɺࣝผਫ਼౓Λେ෯ʹ޲্͍ͤͯ͞Δ ɾFeature Pyramid Network (FPN)ͱ͸ਂ૚৞ΈࠐΈωοτϫʔ Ϋ͕࣋ͭϚϧνεέʔϧͷϐϥϛουܕ֊૚ߏ଄Λར༻ͯ͠ɺ Θ͔ͣͳίετͰಛ௃ϐϥϛουΛߏங͢Δ Feature Pyramid Networks for Object Detection https://openaccess.thecvf.com/content_cvpr_2017/papers/Lin_Feature_Pyramid_Networks_CVPR_2017_paper.pdf ʢ2017ʣTsung-Yi Lin1,2, Piotr Dolla ́r1, Ross Girshick1, Kaiming He1, Bharath Hariharan1, and Serge Belongie2 1Facebook AI Research (FAIR) 2Cornell University and Cornell Tech 2021/03/24
2. ࣍ʹಡΉ࿦จଓ͖ ɾHe, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: CVPR (2016)