Re: [Devel-spam] Adjustments to gocr arguments

[decoder <de...@own-hero.net>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Loren Wilton wrote:
>> I've noticed that for some images, I get better results with -l 160
>> -d 2
>>
>> Explanation of the params:
>>
>> -l num    - threshold grey level 0<160<=255 (0 = autodetect)
>> -d num    - dust_size (remove smaller clusters, -1 = autodetect)
>
> I did *quite* a lot of experimenting with these values and various
> helper programs like pbmnorm the other day.  -l between 160 and 180
> helps a whole lot of spams get recognized.  -d 2 can help or hurt.
> It is good for upright text, but tends to hurt some of the italic text.
>
> Unfortunately there are some more complex stock gifs that look like
> web pages with different colors of shaded backgrounds that will do
> *much* worse at 160 than the default 128.  Basically everything in
> the darker background disappears.
>
> The only thing I could think of would be to run a gif through with
> (or without) -l 160, and if it didn't scpre as spam, try it the
> other way.  But this requires two runs.
>
>        Loren
>


I've modified my plugin to benefit of all settings, maybe you can pack
some example mails and send them to me? (those that you think that
would be hard to recognize by the current FuzzyOcr).


Chris
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFE5tErJQIKXnJyDxURAo1pAJ9RQ582cexggTSqyQU4ax166OkhzACfcwkW
SgI900jJOkex/8qf+f414bA=
=atYR
-----END PGP SIGNATURE-----

Re: [Devel-spam] Adjustments to gocr arguments

[jdow <jd...@earthlink.net>]

From: "decoder" <de...@own-hero.net>

> decoder wrote:
>> Loren Wilton wrote:
>> >>> I've noticed that for some images, I get better results with -l 160
>> >>> -d 2
>> >>>
>> >>> Explanation of the params:
>> >>>
>> >>> -l num    - threshold grey level 0<160<=255 (0 = autodetect)
>> >>> -d num    - dust_size (remove smaller clusters, -1 = autodetect)
>> >> I did *quite* a lot of experimenting with these values and various
>> >> helper programs like pbmnorm the other day.  -l between 160 and 180
>> >> helps a whole lot of spams get recognized.  -d 2 can help or hurt.
>> >> It is good for upright text, but tends to hurt some of the italic text.
>> >>
>> >> Unfortunately there are some more complex stock gifs that look like
>> >> web pages with different colors of shaded backgrounds that will do
>> >> *much* worse at 160 than the default 128.  Basically everything in
>> >> the darker background disappears.
>> >>
>> >> The only thing I could think of would be to run a gif through with
>> >> (or without) -l 160, and if it didn't scpre as spam, try it the
>> >> other way.  But this requires two runs.
>> >>
>> >>        Loren
>> >>
>>
>>
>> I've modified my plugin to benefit of all settings, maybe you can pack
>> some example mails and send them to me? (those that you think that
>> would be hard to recognize by the current FuzzyOcr).
>>
>>
>> Chris
> 
> Now that went to the wrong ML ;) please ignore kthx ;D

If I knew where he had the pictures I'd box them up and ship them
to you. But the FedEx store is closed. (Actually he seems to be
hors du stomachache at the moment.) Would a zip archive of the
images work for you?

{^_^}   Joanne

Re: [Devel-spam] Adjustments to gocr arguments

[decoder <de...@own-hero.net>]

decoder wrote:
> Loren Wilton wrote:
> >>> I've noticed that for some images, I get better results with -l 160
> >>> -d 2
> >>>
> >>> Explanation of the params:
> >>>
> >>> -l num    - threshold grey level 0<160<=255 (0 = autodetect)
> >>> -d num    - dust_size (remove smaller clusters, -1 = autodetect)
> >> I did *quite* a lot of experimenting with these values and various
> >> helper programs like pbmnorm the other day.  -l between 160 and 180
> >> helps a whole lot of spams get recognized.  -d 2 can help or hurt.
> >> It is good for upright text, but tends to hurt some of the italic text.
> >>
> >> Unfortunately there are some more complex stock gifs that look like
> >> web pages with different colors of shaded backgrounds that will do
> >> *much* worse at 160 than the default 128.  Basically everything in
> >> the darker background disappears.
> >>
> >> The only thing I could think of would be to run a gif through with
> >> (or without) -l 160, and if it didn't scpre as spam, try it the
> >> other way.  But this requires two runs.
> >>
> >>        Loren
> >>
>
>
> I've modified my plugin to benefit of all settings, maybe you can pack
> some example mails and send them to me? (those that you think that
> would be hard to recognize by the current FuzzyOcr).
>
>
> Chris

Now that went to the wrong ML ;) please ignore kthx ;D