how to save as picture either or xml about p2pala

[lquirosd]

PAGE complaint XML file is saved by default in <work_dir>/results/<test,val,prod>/page/, to visualize the results I recommend Traskribus or nw-page-editor. Please see USAGE section on main README file, point 4.

[ziyuwzf]

how to save as picture？can you give me some advice？

[lquirosd]

Since document layout could be very complex I don't recommend to save it as a picture directly, but if you want just for baselines there is a script drawBaselines you can compile stand alone using g++ -o drawBaselines drawBaselines.cc -lopencv_imgproc -lopencv_core -lopencv_highgui -lpugixml ( libpugixml and opencv must be available in your system sudo apt-get install libpugixml-dev libopencv-dev)

[longwall]

Thanks for your comment!
In fact I understood zero-size unix symlinks on Windows last night..
I managed to visualize the resulting contours with the simple modification of gen_page() .

Added inside the l_cont loop :

cv2.drawContours(out_img, [l_cnt], 0, (0, 0, 255), 3)

I commented out symlinks generation and added after the .save_page()

cv2.imwrite(os.path.realpath( os.path.join(out_folder, img_name)), out_img)
where out_img is a copy of o_img.

That demonstrates red polygons around htr lines and it's quite enough for my purposes.
By the way - do you plan to publish any other models except ALAR?
Would be interesting to see.
In many cases the actual model can't recognise close verrtically adjacent lines. But generally the algorythm looks cool!