uuencode-generated text is OCRed with many mistakes
yurivict opened this issue · comments
Current Behavior
This text which was generated by the UNIX uuencode command:
is OCRed incorrectly:
M=N[-?3NGO6HU7W; 7SF#W; INU]6\UXMU<@:4]NKJ5Q -#I5>KO0+:!>=[UUZ73
MVTV*-0=[TU=[W>]Q[=U!3WMPS5M[#VO[ 'KWO/>NA>G%1T$*>SW=Z\P]=Y%T]"
M8-$(( )B9 )A,3333(:INI@TT#3333%#0,F0::33TT ">@T:4])Z8)HTP)A,
Tesseract added spaces which aren't present, and failed to detect clearly visible back-quotes, among other issues.
Versions: tesseract-5.3.4, tesseract-data-4.1.0
FreeBSD 14.0
Expected Behavior
n/a
Suggested Fix
n/a
tesseract -v
tesseract 5.3.4
leptonica-1.82.0
libgif 5.2.1 : libjpeg 8d (libjpeg-turbo 3.0.1) : libpng 1.6.40 : libtiff 4.4.0 : zlib 1.3 : libwebp 1.3.2
Found SSE4.1
Found OpenMP 201811
Found libarchive 3.7.2 zlib/1.3 liblzma/5.4.4 bz2lib/1.0.8 liblz4/1.9.4 libzstd/1.5.5
Found libcurl/8.5.0 OpenSSL/3.0.12 zlib/1.3 libpsl/0.21.2 (+libidn2/2.3.4) libssh2/1.11.0 nghttp2/1.58.0
Operating System
No response
Other Operating System
FreeBSD 14.0
uname -a
FreeBSD xx.xx.xx 14.0-STABLE FreeBSD 14.0-STABLE #1 stable/14-n266076-2001d7f6a272: Sat Dec 30 13:33:21 PST 2023 unknown@nohost.com:/disk-samsung/obj/disk-samsung/freebsd-src/amd64.amd64/sys/GENERIC amd64
Compiler
No response
CPU
n/a
Virtualization / Containers
n/a
Other Information
n/a
With -l Latin, the result includes backquotes:
M=N [ - ?3NGO6HU7W ; 7ZSF#W; INU] 6\UXMU<@:4]NKJ5Q`-#I5>K0+: !>=[UUZ73
MVTV* -0=[TU=[W>]Q[=U!3WMP5M[#V9 ['KW0O/>NA>G%1T$*>SW=Z\P]=Y%T] `
M8-$((`)B9`)A,3333(:I^I@QTT#3333$#0,F0::33TT`">@T:4])Z8)HTP)A,
The recognition quality is not a Tesseract issue. It depends on the neural network which is used. In this case most models were trained with texts and languages which require a space after a comma, for example, so it is expected that such models will add such spaces.
If you want to decode uuencoded text, training such data would help. Or try to whitelist possible and to blacklist unexpected characters.