Wikitude

Recognising Hardware components!

Hi Nicolas,

  Thanks for your information and i apologizes for the late reply.

Due to hard time @ my project i couldnt get time to reply.

We've placed some image stickers on the hardwares and able to detect them :)

Basically, those stickers hold asset id.So, it would there for all our componenets.

I've some basic querries about wikitude IR features,

1) Can i detect multiple images using wikitude?

2) Does wikitude support image to text?

 

Thanks

Karthikeyan.B

 

Hello Karthikeyan.B,

 

1) Can i detect multiple images using wikitude?

You can add up to 1000 images in a target collection and detect any of them without network connection, however only one image will be detected in the camera frame at any given time at the moment.

 

2) Does wikitude support image to text?

That is currently not possible, however I will take note of this suggested feature so it may appear in a future version of the SDK.

 

What is your particular use case that makes use of multiple concurrent augmentations and OCR, maybe there is an alterantive approach to it?

 

Best regards,

Nicoloas

 

 

Hi Nicoloas,

Thanks for your information.

 

We are developing an app which recognition the road signs n based on the road sign it tells the user like "Left turn ahead".

In this scenario, there might be one or more signs at a time.In some cases , there might be a specific distance on the image, we just want convey that message to the user.

 

Thanks,

Karthikeyan.B

Hi Karthikeyan.B,

 

thanks for your feedback.

I am sorry to say that for the described use case there does not seem to be a good workaround, indeed. However I took note of the two requested features for internal discussion so they may appear in a future release of the Wikitude SDK.

 

Best regards,

Nicolas

Hi ,

 

We are working on an app which recognize the machineries.

Here we faces an issue with identifing the images perfectly.

For example, how could we recogize the desktop monitors?

Each time, there will be  a different image on the monitor, so we are analysing to use the 3D models.

Can we upload the 3D models on the Tracker tool or vuforia prodives such a facility?

 

Thanks in Advance.

 

 

Hi,

the Wikitude SDK supports detecting and augmenting planar targets, it is currently not possible to detect an arbitrarily shaped 3d object.

However for many use cases, this kind of functionality can be approximated by using multiple planar targets that represent the object, i.e. use images for front, top and side views of the monitor.

In any case the "contents" of the screen must be masked out, for example by using a white area for the screen. That tells the image recognition algorithms to look only for features of the monitor frame and ignore any infomration that is actually displayed. However you have to make sure that the monitor frame contains enough image features with texture and local contrast so the algorithms are able to detect it. 

All the best,

Nicolas

 

Hi Nicolas,

 

Thanks for your infromation.

We just captured the monitor and just masked out the screen with white screen. But, it couldn't recognize it. 

Is there any specific format for the image?

 

Hi Karthikeyan,

what is the star rating of the resulting target image with the white rectangle?

If it is rated with 0 stars it means there are too few image features to reliably detect them.

Best regards,

Nicolas