Google AutoML - detect form fields on an image - google-vision

Is there a way to get coordinates from an form field on an image (scanned image), by using Google vision?
With the (LocalizedObjectAnnotation) can Google detect only objects and creatures
Google OCR (fullTextAnnotation) detects only text
Scenario:
I got an scanned formular. From this scan i would get all form field-positions (input-fields).
It don't work with one or both google method's "LocalizedObjectAnnotation" and "fullTextAnnotation". Because one detect only objects / creatures and the other one only text. So both can't find the input-field in the image.
Has anyone an idee how i get the coordinates for the input-fields?

Related

how to pass multiple values for imagesize in google custom search api

I am using google custom search API to show images, now I want to display both xxlarge, huge images in search results.
Custom search api link:
https://www.googleapis.com/customsearch/v1?key={your Key}&cx={Search
Engine
Id}&q=hello&searchType=image&alt=json&safe=high&filter=1&num=10&fileType=jpg&
imgSize=huge&start=1
In the above link for image size currently I can get huge images, I want to include xxlarge images.
I have the same issue, there doesn't seem to be a way of specifying multiple image sizes, or indeed or doing larger than a certain size, as you can do with google search tools (i.e. on google.com).
so what I am thinking of doing to work around this is calling 2 searches, and adding both results that I want to an array to combine them.

DSC-HX400 RAW image data & Movie Recording

I am currently testing a DSC-HX400. While I am able to do almost everything I need to with the camera there are a couple of items that are not exposed via the API that have frustrated my efforts.
1) The camera does not seem to offer an option, via the API or the camera itself, to capture images in RAW format. It does offer standard & fine JPEG format but both of those are leaving artifacts in the image that become extremely noticeable when you zoom in with an image editor. Is there a way to get the camera to capture RAW images? I do not need the SDK to return the data just to save it out to the card. If getting the RAW data is impossible has anyone found an inventive way to clean up the artifacts?
2) The camera supports both still shoot and movie mode but the API will only expose the mode that I am currently in. It makes it impossible to transition between still to movie mode (to allow recording) from the API but I can do that same transition by pressing a single button on the camera. Once I am recording a movie the API will allow me to transition back to still mode (by cancelling recording). Is there plans to support the ability to trigger a movie recording via the API if you are in a still capture mode (Seeing the firmware already supports this functionality)?
Answers to the questions below:
If the camera cannot capture RAW images, the API will not be able to either. I do not know of a way to capture RAW images but can only comment with regards to the API as I am not an expert on usage of the camera itself.
You can change between still and movie mode by using the "setShootMode" API.

Incorrect image displayed when Gigya opens Google Plus share modal

I'm working with Gigya's Share Bar javascript api and am having an issue with google plus sharing. I have separate UserAction objects for facebook, twitter and google plus to display different text and linkbacks. I'm adding a media item (image) to the facebook and google plus actions to share that image along with the text. Facebook is sharing the correct image, however, google plus is displaying a different image. I have no idea where this image is coming from. Text and link are correct on both, but the image is frustratingly incorrect.
Has anyone encountered an issue like this? Is there something different that must be done for google plus shares?
Test your target URL with Google's structured data testing tool.
Read the snippet documentation to understand how the Google crawler determines what to use in the share preview.

generate geo tiles from vector data

i am facing this problem : I have in my database some vector data (polylines ...) which describe many world/country/region/cities boundaries.
I would like to have an interactive map on thoses areas.
When i render those data in the browser (data are sent in geojson) this is slow (i'am not surprised).
The next reason, is to generate some raster tiles to represent the areas at different zoom level and to have a way to make the relation between a click on a area and the area selected (to display more info).
I am a newbie in geo tiles :
What is the best tool to generate png/jpg tiles from vector data?
How to make relation between an area clicked and a "id" in database?
I know my questions are very large, but if you have some advices ...
Thanks!
I recently wrote a blog posting on how to improve the rendering speed of raster tiles:
http://blog.webfoot.com/2013/03/12/optimizing-map-tile-generation/
As for connecting a click to a jurisdiction ID, I don't think there is a shortcut: you need to translate the click into lat/lng (or Vast Coordinate System, as I describe in the blog post) and then do a point-in-polygon check. If you are using PostGIS, then you can do an ST_Intersects.
NOTE: if you are using MySQL, the intersection is done with bounding boxes, not the polygons. It still is possible: you just need to grab a point-in-polygon function from somewhere to check which polygon if a bounding-box check gets you more than one.
Here is an example of dots you can click on:
http://maps.webfoot.com/demos/CanadianStimulus/CanadianStimulus.html
Note that because people aren't precise with clicking (especially with tiny dots), I actually make a few pixels around the dots "live" as well as the single-pixel dots. You can also click away from the dot, and then it will show the polygon (which in this example's case, is the federal electoral district).
There is a MongoDB Plugin datasource for Geoserver, a full featured WebGIS Server. You can use it to build up a map rendering service (a WMS actually) ofyour data. Then use Geowebcache to generate tiles of your map.
This way you can use fully featured WebGIS stack, which allows you to manage additional datasources, and easy tile subset recreation if portions of your data get updated.
Also querying features on a map (relation map position to element in database) is provided through Geoservers WFS implementation.

pdf overlay touch in an app

i'm looking for pdf overlay feature to be implemented in iOS app, if whether its possible with libs free or even paid to have an touch overlay on PDF page within an app?
what i'm trying to do is i have a map which is train map and i want to have the map as pdf and any touch on station brings up info on each station.
is that posssible or what is other way to go forward open to every opinion!!!
i saw some other questions but none directed like this i.e.
Overlay Image above Embedded pdf
can any one tell me if its actually possible in objctive C code?
needs to work as offline app.
You can use Google Maps Static API (https://developers.google.com/maps/documentation/staticmaps/) to return an image on the map and then convert it to pdf.