Guide for sane support for Avision FB2280E
Hello
Update 21/3 2019. Only color mode are supported. Not gray or lineart. scanimage gives sane_start: Error during device I/O mode. And I have to unplug power on scanner.
I picked up my new Avision FB2280E yesterday. I had googled a lot before and had hope it should be usable under linux. It can use the same driver as FB2080E but for some unknown reason it's not enabled in sane.
So this is a note to self, that maybe others find usefull.
It's based on this pages.
https://forum.ubuntuusers.de/topic/u...nicht-erkannt/
https://tovotu.de/blog/546-Buchkante...gitalisierung/
First we install some tools and download sane-backends
Code:
mkdir ~/sane
cd ~/sane
sudo apt-get install autopoint autogen libtool shtool autoconf-archive
sudo apt-get install libusb-dev
sudo apt install git
git clone https://gitlab.com/sane-project/backends.git
cd backends
Then I create a patch for avision.c
Code:
echo " " |tee avision-patch
echo " { NULL, NULL," |tee -a avision-patch
echo " 0x0638, 0x2a1f," | tee -a avision-patch
echo " \"Avision\", \"FB2280E\"," | tee -a avision-patch
echo " 0}," | tee -a avision-patch
echo " /* comment=\"1 pass, 600 dpi, zero-edge\" ASIC 7 */" | tee -a avision-patch
echo " " |tee -a avision-patch
Next step are a little hack. I find a linenumber just above the FB2080E
and insert the patch there.
Code:
line=`cat -n backend/avision.c| grep -B 3 FB2080 |head -n1|tr -d ' '`
insert above patch at linenumber(be carefull here)
printf "%s\n" "${line}r avision-patch" w |ed -s backend/avision.c
Time to built. Look out for error on missing dependencies.
Code:
./autogen.sh
./configure --prefix=/usr --sysconfdir=/etc --localstatedir=/var
make
sudo make install
The libearies libsane-avi* needs to be accessible by sane. I just copy all of them.
Code:
sudo cp -v -a /usr/lib/sane/* /usr/lib/x86_64-linux-gnu/sane/
Check that they are linking correct. The should now link like this (there are a few more files here)
Code:
ls -l /usr/lib/x86_64-linux-gnu/sane/libsane-avi*
/usr/lib/x86_64-linux-gnu/sane/libsane-avision.so.1 -> libsane-avision.so.1.0.29
#check also libsane-dll*
ls -l /usr/lib/x86_64-linux-gnu/sane/libsane-dll*
/usr/lib/x86_64-linux-gnu/sane/libsane-dll.so -> libsane-dll.so.1.0.29
Test scanimage -V
Code:
scanimage -V
scanimage (sane-backends) 1.0.29-293-g1ed4c7d8-dirty; backend version 1.0.29
Now I was able to scan as sudo user!
Code:
sudo scanimage -d 'avision' --format pnm > out.pnm
But, not as normal user. I add a few more steps.
Making a new patch. I'm still in the same folder.
Code:
echo '' | tee patch.txt
echo ':model "FB2280E"' | tee -a patch.txt
echo ':interface "USB"' | tee -a patch.txt
echo ':usbid "0x0638" "0x2a1f"' | tee -a patch.txt
echo ':comment "1 pass, 600 dpi, zero-edge" ASIC 7' | tee -a patch.txt
echo ':status :basic' | tee -a patch.txt
Patching
Code:
sed -i '/:mfg "Avision"/ r patch.txt' doc/descriptions/avision.desc
Generating hwdb file
Code:
tools/sane-desc -s doc/descriptions -m hwdb |sudo tee /lib/udev/hwdb.d/20-sane.hwdb
Creating udev-rules
Code:
echo 'ATTRS{manufacturer}=="AVISION", DRIVERS=="usb", SUBSYSTEM=="usb", ATTRS{idVendor}=="0638", ATTRS{idProduct}=="2a1f", ENV{libsane_matched}="yes", MODE="0777"' | sudo tee -a /etc/udev/rules.d/55-udev-avision.rules
Some add the device to /etc/sane.d/avision.conf
#usb 0x0638 0x2a1f
Don’t do that. it will give a warning like this when running scanimage
[avision] attach: "AVISION" - "FB2280E" not yet in whitelist!”
#reboot or
Code:
sudo udevadm control --reload-rules
sudo udevadm trigger
And we can now scan :-)
Code:
scanimage -d 'avision' --format pnm > out.pnm
At least until next update.
I had some problems when I tried to figure out this. I set this debug variable, and it was great help.
Code:
export SANE_DEBUG_DLL=255
I have not tested the scanner much yet. I just had to document everything when it was fresh in mind. Hope someone find this usefull.
Re: Guide for sane support for Avision FB2280E
I primarily bought this scanner for book scanning. But I had to test the photo capabilities.Straight out of the scanner the photos are washed out. Fortunately have I a scan target and a spectrometer somewhere....
Output from command
Code:
scanimage -d 'avision' --resolution 600 --format tiff --mode "16bit Color" | convert - -rotate 90 albumsheet.jpg
https://drive.google.com/thumbnail?a...4oshL1Hfp3p_fN
Full size
After scanning the target and creating a profile with argyllcms, I got this output. No other processing applied.
Code:
scanimage -d 'avision' --resolution 600 --format tiff --mode "16bit Color" -i Avision_FB2280E.icc | convert - -intent Relative -black-point-compensation -profile /usr/share/color/icc/colord/sRGB.icc -rotate 90 albumsheet-color-corrected.jpg
https://drive.google.com/thumbnail?a...tVtqrgTapp0I9M
Full size
This color profile may work with FB2080E also.
You can find it here.
Re: Guide for sane support for Avision FB2280E
Scanning time
Just some more info:
Scanning time in seconds for color scan:
Code:
Resolution Full page Novel page
100 3.42 3.12
200 3.44 3.12
300 3.46 3.11
301 9.78 7.77
350 9.79 7.78
400 9.82 7.78
500 9.85 7.78
600 9.87 7.78
Scanned quality
I scanned 11 page from a novel where I found a sample chapter at the publisher. Then I was able to compare the quality. For the quality control I used ocrevalutf8 from https://github.com/ryanfb/ancientgre...aluation-tools I created my groundtruth file like this: It was a two colum paged file.
Code:
k2pdfopt -mode 2col sample_chapter.pdf
~/Downloads/xpdf-tools-linux-4.02/bin64/pdftotext -layout -enc UTF-8 sample_chapter_k2opt.pdf - > groundtruth.txt
Accuracy check was done like this:
Code:
ocrevalutf8 accuracy groundtruth.txt 100pbm.txt |head -n 5
ocrevalutf8 wordacc groundtruth.txt 100pbm.txt |head -n 5
Code:
Reolution UT* TT* pbm pgm ppm
Accuracy/
A/W A/W
Wordaccuracy
100 3.2 6/5.5/6.3 98.44/ 94.44 99.35/ 98.52 99.29/ 98.23
200 11.2 9.6/10/13.2 99.48/ 99.06 99.49/ 99.09 99.47/ 99.06
300 23.2 10.7/13.1/18.6 99.47/ 99.02 99.49/ 99.14 99.49/ 99.14
400 38.7 12.5/16.4/27.9 99.48/ 99.11 99.50/ 99.16 99.58/ 99.16
500 58 14.8/21.5/35.6 99.48/ 99.14 99.49/ 99.11 99.48/ 99.14
600 87.7 18/26.2/44 99.48/ 99.16 99.48/ 99.16 99.48 / 99.11
UT. Average unpaper processing time all 11 sheets from a novel. Ppm time was the fastest, but newer more than 1sec faster.
TT. Average tesseract processing time
I'm very impressed by the quality of tesseract.
My full processing workflow was like this:
#unpaper step
Code:
$ls -d */
$100/ 200/ 300/ 400/ 500/ 600/
$ls 100/
page006.tif page008.tif page010.tif page012.tif page014.tif page016.tif
page007.tif page009.tif page011.tif page013.tif page015.tif page017.tif
mkdir 100/pbm 100/pgm 100/ppm
#Next line 3 times for every resolution, pbm,pgm and ppm
alias unptime="unpaper -t pgm -start 6 100/page%03d.tif 100/pgm/unpaper%03d.pgm"
time unptime
Ocr was done like this:
Code:
ls 100/pbm/un* > files
alias tesstt="tseseract -l nor --dpi 100 files 100pbm txt"
time tesstt
Just a little update. Quite exited with the results from tseseract I had to test it against a couple of competitors. I have access to Adobe Acrobat DC (not pro) at work. I only testedt the 300 dpi pgm source.
The results where
accuracy 98.05%
wordacc 95.18%
I also found out there was somthing called Azure Cognitive Services from microsoft with Computer Vision API , it was free for one year. The problem was it outputed json file, and not one of the knowen ocr formats. So I hacked together something. It's very basic and only ment for my whises to convert scanned novel pages. The script extract all the json lines boundingBoxes and creates a fig file (xfig package). This file are converted to pdf and text are extracted. Very basic, and use with care.
Scrip as txt or zip.
Anyway results are:
accuracy 99.10% Accuracy
wordacc 99.10% Accuracy
Re: Guide for sane support for Avision FB2280E
Just a simple video describing the scanning process. Sorry for missing English.
https://www.youtube.com/watch?v=15qR8KUzVa4&
