Output previously would have maintained the image size of the included source image or at least made it fit to a page. This no longer happens and the image is a lot smaller.
Cloud OCR output is a lot smaller than expected
この記事は役に立ちましたか？
0人中0人がこの記事が役に立ったと言っています
コメント
9件のコメント
Please try not to use deskew correction, i.e. use correctSkew=false for the processImage method. Using of imageSource=auto (it is by default) is enough to correct some geometrical distortion from your image. I use the following URL:
string url = "http://cloud.ocrsdk.com/processImage?correctSkew=false&exportFormat=pdfSearchable";
And in this case the output is nice.
This has worked previously without specifying this option. Also it appears to shrink the output randomly for similar sources - i.e. taking a photo of pages from the same textbook, some are as expected whereas some are very small.
Has there been any changes on the service side which now require this setting to be specified?
Really we updated the used OCR technologies for the service a few weeks ago, but I have just tested your image with the previous version of OCRT and the result is the same. Thus, this is a particularity of this image. Anyway, I have reported a reclamation to our developers and will let you know about any news as soon as I get an answer.
The developers say that the result is expected, because it is designed so, that during deskew correction we fill the pieces on the sides of the image with the average color of the background. Thus, it is possible that they can be brighter then the image background itself.
About cropping the image when we use correctSkew=false, I will answer you later.
Still having this problem - more and more users are reporting the issue. Have you any solution? I have provided more examples of files it has occurred randomly with (does not happen all the time).
Thank you for the provided images, I have reproduced the issue and seems it is another issue. I'm going to consult with the developers. I will let your know about their answer.
Is there any update about this issue?
Unfortunately, not, the developers are still investigating the issue.
Hello Neil,
We apologize for the long silence and all possible inconvenience. The issue fortunately is fixed and it is available in Cloud OCR SDK beginning from this morning. Could you please test the fix and give us your feedback? Is everything fine now?
サインインしてコメントを残してください。