El 15 abr 2019, a las 7:21, Robert Hänggi <aarjay.robert@gmail.com> escribió:

On 14/04/2019, Noelia Ruiz <nrm1977@gmail.com> wrote:
Hi, making another suggestion: I would like that the used plugin was
mentioned at least in the title of browseable message or document
result. This is useful when we copy some article with graphics and we
want to include automatic descriptions with reference to the origin.

I don't like that as title, but you could append it to the clipboard
result, something like:
"Powered by OnlineImageDescriber".
But you might need to mention the used engine as well.

Robert

Cheers

El 13/04/2019 a las 15:20, Larry Wang escribió:
- a checkbox for OCR if short text should be spoken or always displayed.
'Use browse-able message for text result' is provided for this purpose.
- Always Browse-able Message for image description but eventually
replaced by a faked recognition object if any object positions are
known.
There is faked object result for image description.
If you use the faked object in Azure analyzer position info is included.
I set the faked position of general description to left right corner of
the image. What comes after description is the name of celebrities
landmarks and faces, location information are attached to object name.

Roughly 20% of ocr space result is timeout. May be 10 second timeout is
still not enough for OCR?

As for GUI issue. May be I should put common options on top of engine
settings for convenience.

On 2019/4/13 16:11, Robert Hänggi wrote:
ized objects/faces but I don't see that
here.

----- Original Message -----

From: marcio via Groups.Io

To: nvda-addons@nvda-addons.groups.io

Sent: Monday, April 15, 2019 12:07 AM

Subject: Re: [nvda-addons] different pitch and speed for the keyboard then the general voice

Hi,
I didn't get your idea. Would you mind trying and explaining it again?
Everything that deals with typing always catch my attention :)

Cheers,
Marcio AKA Starboy
Follow or add me on Facebook

Em 15/04/2019 05:49, Isaac disse:

Hi, is there a way to set the typing keyboard to super fast.

Thanks

Hi, is there a way to set the typing keyboard to super fast.

Thanks

- a checkbox for OCR if short text should be spoken or always displayed.
'Use browse-able message for text result' is provided for this purpose.
- Always Browse-able Message for image description but eventually
replaced by a faked recognition object if any object positions are
known.
There is faked object result for image description.
If you use the faked object in Azure analyzer position info is included.
I set the faked position of general description to left right corner
of the image. What comes after description is the name of celebrities
landmarks and faces, location information are attached to object name.

Roughly 20% of ocr space result is timeout. May be 10 second timeout
is still not enough for OCR?

As for GUI issue. May be I should put common options on top of engine
settings for convenience.

On 2019/4/13 16:11, Robert Hänggi wrote:

ized objects/faces but I don't see that here.

- a checkbox for OCR if short text should be spoken or always displayed.
'Use browse-able message for text result' is provided for this purpose.
- Always Browse-able Message for image description but eventually
replaced by a faked recognition object if any object positions are
known.
There is faked object result for image description.
If you use the faked object in Azure analyzer position info is included. I set the faked position of general description to left right corner of the image. What comes after description is the name of celebrities landmarks and faces, location information are attached to object name.
Roughly 20% of ocr space result is timeout. May be 10 second timeout is still not enough for OCR?
As for GUI issue. May be I should put common options on top of engine settings for convenience.
On 2019/4/13 16:11, Robert Hänggi wrote:

ized objects/faces but I don't see that
here.

ized objects/faces but I don't see that
here.

Hi, I agree with most of this. I think that Machine learning could work sometimes, but many tomes doesn't work here.

5. 'Use browseable message for text result' can be changed to 'Show recognition result in browse mode'.

Maybe better to use browseable message, since when using Win20 OCR style, the result is also shown in browse mode, that is, both browseable message and OCR result are shown in different browse mode windows.
Personally, I would remove the possibility of using just ui.message, since generally this is not confortable for braille with long messages like that.

Cheers



El 12/04/2019 a las 21:22, Andre Fisher escribió:

Hi.

A few comments and interesting observations.

1. On my end, the Machine Learning Engine by Oliver Edholm option doesn't work at all. This is also bothersome, as it is the default option.
2. The acronym 'API' should always bee in uppercase, so the option 'Use public api quota' is incorrect.
3. Change 'Max' to 'Maximum' in the number of candidates to avoid confusion.
4. Change option from 'Copy result to clipboard after recognition' to 'Copy recognition result to the clipboard'
5. 'Use browseable message for text result' can be changed to 'Show recognition result in browse mode'.
6. When the above option is unchecked, and the mode used to show the result is similar to the one using Windows 10 OCR, navigation by line, word and paragraph is impossible.
7. The message shown when the option is checked has no title, but appears as 'NVDA message'
8. The option 'Swap the effect of repeated gesture with none repeated ones' isn't clear by it's name. I suggest removing it (see next point why).
9. There is a bug with the add-on. While copying to the clipboard works, there seems to be an issue with 'Swap the effect of repeated gesture with none repeated ones'. (option B) and 'Use browseable message for text result' (option A).
When Option A is checked, and Option B is both checked and unchecked, a browsable message does appear.
When option A is unchecked, and option B is both checked unchecked, the message appears in a Virtual document.
So, option B seems useless, as even with it unchecked, the result still shows in a browse mode/virtual document.
10. 'Enable more verbose logging for debug purpose' should be '...purposes'

Hi.

A few comments and interesting observations.

1. On my end, the Machine Learning Engine by Oliver Edholm option doesn't work at all. This is also bothersome, as it is the default option.
2. The acronym 'API' should always bee in uppercase, so the option 'Use public api quota' is incorrect.
3. Change 'Max' to 'Maximum' in the number of candidates to avoid confusion.
4. Change option from 'Copy result to clipboard after recognition' to 'Copy recognition result to the clipboard'
5. 'Use browseable message for text result' can be changed to 'Show recognition result in browse mode'.
6. When the above option is unchecked, and the mode used to show the result is similar to the one using Windows 10 OCR, navigation by line, word and paragraph is impossible.
7. The message shown when the option is checked has no title, but appears as 'NVDA message'
8. The option 'Swap the effect of repeated gesture with none repeated ones' isn't clear by it's name. I suggest removing it (see next point why).
9. There is a bug with the add-on. While copying to the clipboard works, there seems to be an issue with 'Swap the effect of repeated gesture with none repeated ones'. (option B) and 'Use browseable message for text result' (option A).
When Option A is checked, and Option B is both checked and unchecked, a browsable message does appear.
When option A is unchecked, and option B is both checked unchecked, the message appears in a Virtual document.
So, option B seems useless, as even with it unchecked, the result still shows in a browse mode/virtual document.
10. 'Enable more verbose logging for debug purpose' should be '...purposes'

On 13/04/2019, Noelia Ruiz <nrm1977@gmail.com> wrote:

Robert's approach sounds good, at least regarding messages
About the dialog, for me both approaches, the current one and Robert
suggestions, are acceptable, since API keys deppend on each plugin, so
maybe someone uses an own api key for just the most frequently used
plugin, but if it needs to be changed, the api key would need to be
modified too, so, at least in some cases, api key option would be as
frequent like plugins choices.
Personally, I would build the dialog as follows:
- A readonly for OCR with the changer button.
- The readonly for Image describer and the changer button.

Why can't this be a combo box?
The "Choose" or rather "Details" button would then open the settings
for that particular engine.
One important point is the scope:
What should be changeable per profile and what settings should be general.
I think the access methods could be global (i.e. what is in the dialog
popping up), what do you think?
On the other hand, scaling settings, the engine engaged and the detail
selection could be per profile and should thus stay in the main panel
(makes it necessary to change the content dynamically, depending on
the engine selected).
By the way, I would be glad if the Windows OCR could be selectable as
well, it could profit from extended image filtering, e.g. scaling with
sharpening, converting to grey scale, etc. This might yield better
results.
Robert

- This two main buttons would open dialogs with combo box to select,
respectively, the OCR with their options, including api key for each one
indepently, and the same for image describers. This would have the
advantage that, if other plugins are added in future, options could be
easily customized for each one.

Cheers

El 13/04/2019 a las 10:11, Robert Hänggi escribió:

On 13/04/2019, Noelia Ruiz <nrm1977@gmail.com> wrote:

Hi, I agree with most of this. I think that Machine learning could work
sometimes, but many tomes doesn't work here.

5. 'Use browseable message for text result' can be changed to 'Show
recognition result in browse mode'.

Maybe better to use browseable message, since when using Win20 OCR
style, the result is also shown in browse mode, that is, both browseable
message and OCR result are shown in different browse mode windows.
Personally, I would remove the possibility of using just ui.message,
since generally this is not confortable for braille with long messages
like that.

How about limiting ui.message output to one-liners and only for OCR?

I wouldn't use browseable message either.
The recognition object would be fine if you replace the Windows line
breaks (\CR\LF) with normal ones. It might be necessary to tweak the
text infos because there are no positions for a description result, in
other words, you would have to fake them.
The reason behind using the recognition object is that at some point
the recognized objects and faces might have a position assigned to
them.
If it was a button (e.g. a compact disc), it could perform an action.
You can't achieve that with a browseable message (Sorry for the typo
but I use browseable as it is used by NVDA).
By the way, it is documented in the add-on help that Azure can deliver
the approx position of recognized objects/faces but I don't see that
here.

In summary:
- a checkbox for OCR if short text should be spoken or always displayed.
- Always Browseable Message for image description but eventually
replaced by a faked recognition object if any object positions are
known.

Oliver's engine works here but OCR Space gives me 99 % of the time a time
out.

Also, I would turn the options in the GUI around such that the ones
that are frequently changed are on the main panel.
That is, the engine should be a combo box, not a read-only field
whereas the API access details should be hidden in a sub-menu (button
next to the combo) since they should work once they are set up.

Cheers
Robert

Cheers



El 12/04/2019 a las 21:22, Andre Fisher escribió:

Hi.

A few comments and interesting observations.

1. On my end, the Machine Learning Engine by Oliver Edholm option
doesn't work at all. This is also bothersome, as it is the default
option.
2. The acronym 'API' should always bee in uppercase, so the option 'Use
public api quota' is incorrect.
3. Change 'Max' to 'Maximum' in the number of candidates to avoid
confusion.
4. Change option from 'Copy result to clipboard after recognition' to
'Copy recognition result to the clipboard'
5. 'Use browseable message for text result' can be changed to 'Show
recognition result in browse mode'.
6. When the above option is unchecked, and the mode used to show the
result is similar to the one using Windows 10 OCR, navigation by line,
word and paragraph is impossible.
7. The message shown when the option is checked has no title, but
appears as 'NVDA message'
8. The option 'Swap the effect of repeated gesture with none repeated
ones' isn't clear by it's name. I suggest removing it (see next point
why).
9. There is a bug with the add-on. While copying to the clipboard
works,
there seems to be an issue with 'Swap the effect of repeated gesture
with none repeated ones'. (option B) and 'Use browseable message for
text result' (option A).
When Option A is checked, and Option B is both checked and unchecked, a
browsable message does appear.
When option A is unchecked, and option B is both checked unchecked, the
message appears in a Virtual document.
So, option B seems useless, as even with it unchecked, the result still
shows in a browse mode/virtual document.
10. 'Enable more verbose logging for debug purpose' should be
'...purposes'

Robert's approach sounds good, at least regarding messages
About the dialog, for me both approaches, the current one and Robert suggestions, are acceptable, since API keys deppend on each plugin, so maybe someone uses an own api key for just the most frequently used plugin, but if it needs to be changed, the api key would need to be modified too, so, at least in some cases, api key option would be as frequent like plugins choices.
Personally, I would build the dialog as follows:
- A readonly for OCR with the changer button.
- The readonly for Image describer and the changer button.
- This two main buttons would open dialogs with combo box to select, respectively, the OCR with their options, including api key for each one indepently, and the same for image describers. This would have the advantage that, if other plugins are added in future, options could be easily customized for each one.
Cheers
El 13/04/2019 a las 10:11, Robert Hänggi escribió:

On 13/04/2019, Noelia Ruiz <nrm1977@gmail.com> wrote:

Hi, I agree with most of this. I think that Machine learning could work
sometimes, but many tomes doesn't work here.

5. 'Use browseable message for text result' can be changed to 'Show
recognition result in browse mode'.

Maybe better to use browseable message, since when using Win20 OCR
style, the result is also shown in browse mode, that is, both browseable
message and OCR result are shown in different browse mode windows.
Personally, I would remove the possibility of using just ui.message,
since generally this is not confortable for braille with long messages
like that.

How about limiting ui.message output to one-liners and only for OCR?

I wouldn't use browseable message either.
The recognition object would be fine if you replace the Windows line
breaks (\CR\LF) with normal ones. It might be necessary to tweak the
text infos because there are no positions for a description result, in
other words, you would have to fake them.
The reason behind using the recognition object is that at some point
the recognized objects and faces might have a position assigned to
them.
If it was a button (e.g. a compact disc), it could perform an action.
You can't achieve that with a browseable message (Sorry for the typo
but I use browseable as it is used by NVDA).
By the way, it is documented in the add-on help that Azure can deliver
the approx position of recognized objects/faces but I don't see that
here.

In summary:
- a checkbox for OCR if short text should be spoken or always displayed.
- Always Browseable Message for image description but eventually
replaced by a faked recognition object if any object positions are
known.

Oliver's engine works here but OCR Space gives me 99 % of the time a time out.

Also, I would turn the options in the GUI around such that the ones
that are frequently changed are on the main panel.
That is, the engine should be a combo box, not a read-only field
whereas the API access details should be hidden in a sub-menu (button
next to the combo) since they should work once they are set up.

Cheers
Robert

Cheers



El 12/04/2019 a las 21:22, Andre Fisher escribió:

Hi.

A few comments and interesting observations.

1. On my end, the Machine Learning Engine by Oliver Edholm option
doesn't work at all. This is also bothersome, as it is the default
option.
2. The acronym 'API' should always bee in uppercase, so the option 'Use
public api quota' is incorrect.
3. Change 'Max' to 'Maximum' in the number of candidates to avoid
confusion.
4. Change option from 'Copy result to clipboard after recognition' to
'Copy recognition result to the clipboard'
5. 'Use browseable message for text result' can be changed to 'Show
recognition result in browse mode'.
6. When the above option is unchecked, and the mode used to show the
result is similar to the one using Windows 10 OCR, navigation by line,
word and paragraph is impossible.
7. The message shown when the option is checked has no title, but
appears as 'NVDA message'
8. The option 'Swap the effect of repeated gesture with none repeated
ones' isn't clear by it's name. I suggest removing it (see next point
why).
9. There is a bug with the add-on. While copying to the clipboard works,
there seems to be an issue with 'Swap the effect of repeated gesture
with none repeated ones'. (option B) and 'Use browseable message for
text result' (option A).
When Option A is checked, and Option B is both checked and unchecked, a
browsable message does appear.
When option A is unchecked, and option B is both checked unchecked, the
message appears in a Virtual document.
So, option B seems useless, as even with it unchecked, the result still
shows in a browse mode/virtual document.
10. 'Enable more verbose logging for debug purpose' should be
'...purposes'

Hy Nomi and all,

This version is provisional, until the new version comes out, currently
the author is working on his code, correcting errors, etc.

Weather Plus (Apixu) version 5.5 using Adriano's Key):

http://www.nvda.it/files/plugin/weather_plus5.5.nvda-addon

Thanks for using the Weather_Plus add-on!

Cheers.

Rémy


Le 13/04/2019 à 08:46, Nomi a écrit :

Hi dear listers,
Hope you'll all be fine.
Friends I need an addon called weather plus, Actually I searched it on
google, but Provided link was not working and giving error.
So I request if anyone from the list has this add-on, Please Share it
in the group.
Regards

On 13/04/2019, Noelia Ruiz <nrm1977@gmail.com> wrote:

Hi, I agree with most of this. I think that Machine learning could work
sometimes, but many tomes doesn't work here.

5. 'Use browseable message for text result' can be changed to 'Show
recognition result in browse mode'.

Maybe better to use browseable message, since when using Win20 OCR
style, the result is also shown in browse mode, that is, both browseable
message and OCR result are shown in different browse mode windows.
Personally, I would remove the possibility of using just ui.message,
since generally this is not confortable for braille with long messages
like that.

How about limiting ui.message output to one-liners and only for OCR?
I wouldn't use browseable message either.
The recognition object would be fine if you replace the Windows line
breaks (\CR\LF) with normal ones. It might be necessary to tweak the
text infos because there are no positions for a description result, in
other words, you would have to fake them.
The reason behind using the recognition object is that at some point
the recognized objects and faces might have a position assigned to
them.
If it was a button (e.g. a compact disc), it could perform an action.
You can't achieve that with a browseable message (Sorry for the typo
but I use browseable as it is used by NVDA).
By the way, it is documented in the add-on help that Azure can deliver
the approx position of recognized objects/faces but I don't see that
here.
In summary:
- a checkbox for OCR if short text should be spoken or always displayed.
- Always Browseable Message for image description but eventually
replaced by a faked recognition object if any object positions are
known.
Oliver's engine works here but OCR Space gives me 99 % of the time a time out.
Also, I would turn the options in the GUI around such that the ones
that are frequently changed are on the main panel.
That is, the engine should be a combo box, not a read-only field
whereas the API access details should be hidden in a sub-menu (button
next to the combo) since they should work once they are set up.
Cheers
Robert

Cheers



El 12/04/2019 a las 21:22, Andre Fisher escribió:

Hi.

A few comments and interesting observations.

1. On my end, the Machine Learning Engine by Oliver Edholm option
doesn't work at all. This is also bothersome, as it is the default
option.
2. The acronym 'API' should always bee in uppercase, so the option 'Use
public api quota' is incorrect.
3. Change 'Max' to 'Maximum' in the number of candidates to avoid
confusion.
4. Change option from 'Copy result to clipboard after recognition' to
'Copy recognition result to the clipboard'
5. 'Use browseable message for text result' can be changed to 'Show
recognition result in browse mode'.
6. When the above option is unchecked, and the mode used to show the
result is similar to the one using Windows 10 OCR, navigation by line,
word and paragraph is impossible.
7. The message shown when the option is checked has no title, but
appears as 'NVDA message'
8. The option 'Swap the effect of repeated gesture with none repeated
ones' isn't clear by it's name. I suggest removing it (see next point
why).
9. There is a bug with the add-on. While copying to the clipboard works,
there seems to be an issue with 'Swap the effect of repeated gesture
with none repeated ones'. (option B) and 'Use browseable message for
text result' (option A).
When Option A is checked, and Option B is both checked and unchecked, a
browsable message does appear.
When option A is unchecked, and option B is both checked unchecked, the
message appears in a Virtual document.
So, option B seems useless, as even with it unchecked, the result still
shows in a browse mode/virtual document.
10. 'Enable more verbose logging for debug purpose' should be
'...purposes'