Detecting page angle currently requires that automatic page segmentation is enabled. Therefore, if rotateAuto is set to true but the current PSM does not support detecting the page angle, page segmentation is run with PSM set to 3 (AUTO), the page angle is retrieved, and then page segmentation is run a second time with PSM set to whatever is requested by the user.
If PSM is set to 3 by the user already, then page segmentation should only be run once (if no rotation is detected). However, due to a bug in the implementation of this feature, page segmentation is currently being run twice, even in this case.
Specifically, this code does not account for the fact that the PSM object stores the PSM values as strings, but api.GetPageSegMode returns an integer. Therefore, the check that occurs is ['3'].includes(3) which resolves to false.
Detecting page angle currently requires that automatic page segmentation is enabled. Therefore, if
rotateAutois set totruebut the currentPSMdoes not support detecting the page angle, page segmentation is run withPSMset to3(AUTO), the page angle is retrieved, and then page segmentation is run a second time withPSMset to whatever is requested by the user.If
PSMis set to3by the user already, then page segmentation should only be run once (if no rotation is detected). However, due to a bug in the implementation of this feature, page segmentation is currently being run twice, even in this case.https://github.com/naptha/tesseract.js/blob/master/src/worker-script/index.js#L402-L407
Specifically, this code does not account for the fact that the
PSMobject stores the PSM values as strings, butapi.GetPageSegModereturns an integer. Therefore, the check that occurs is['3'].includes(3)which resolves tofalse.