pone.0296314.g003.tif (721.07 kB)
Data augmentation.
figure
posted on 2024-01-05, 18:31 authored by Zhi-Ben Yin, Fu-Yong Liu, Hui Geng, Ya-Jun Xi, De-Bin Zeng, Chun-Jing Si, Ming-Deng Shi(a) Original drawing; (b) Horizontal rotation; (c) Vertical rotation; (d) Random rotation with a rotation angle between 0 and 90 degrees; (e) Random cutting with a pixel value between 0 and 30 pixels; (f) Image scaling compared to the original drawing is between -0.2 and 0.2; (g) Image brightness; (h) Image translation compared to the original drawing is between 0 and ±0.3.
History
Usage metrics
Categories
Keywords
traditional convolution layertarget detail featuressingle shot multimultiple location detailsjujubessd reached 97href =" mailtoexperimental results demonstratedeformable convolutional networksbalanced feature pyramid61 %, respectively1 %, representingparameter information obtainedmethod called jujubessdidentifying disease spotsdisease spot regionsmean average precisiondiv >< pdisease spotsweak informationspot detectionproposed methodvarying sizesvarying densitiesvaried sizesprecise detectionpractical applicationpaper proposesoriginal algorithmobjectives duefaster rexisting algorithmsdcns ).data augmentationcurrent algorithmsbox detectorartificial collectionapproximately 6
Licence
Exports
RefWorks
BibTeX
Ref. manager
Endnote
DataCite
NLM
DC