Log in or register to post comments

Detect Image Target via Hololens Camera

March 5, 2019 - 10:36am #1

Hi guys,

Currently I am quite confuse about the way vuforia detect Image Target.

Take an example below:

I would like to create an AR App for hololens, which can show informations about some specific thing (AR like) in the realworld, which is a 3D Object (eg. a big house) when i look at it.

for example when i use the hololens to view a house. i would like to see the information about that house like size, building time etc...

So like from the basic tutorial example from vuforia (https://assetstore.unity.com/packages/templates/packs/vuforia-core-samples-99026), I would say:

Firstly I need to define a 3d model target of that house via model target generator or maybe Object Scanner???. then use it as 3D Model Target in Unity and when the  hololens camera detected the model target, i can start to show the specific information (eg. size, building time, etc...).

For me the house are like the Image Target which is similar to the cards from vuforia samples (but in 3D so it has to be model target) and the information which need to be show  will be in ar view.

 

How do I suppose to do that ? From my understanding I have to save the House as Target (which one to fulfill the job, i still cant figure out) in vufodia target manager.

I am searching for the right way to realize. Currently I know Model Target Generator and Object Scanning. I dont know if I should use them to do my work .

One more question would be.

Do i have to define the Target manually with vuforia target manager everytime when i want to add a new target to my application, which need to be detected or can I do that dynamically via programming. 

 

Thanks

Detect Image Target via Hololens Camera

March 7, 2019 - 10:34am #3

medabit wrote: If you had a 3D model of the home you could certainly use that for Model Targets, and is likely a good use case. I'm assuming the challenge here is to get a 3D model that can be imported into the MTG, and is accurate enough to be recognized by the Vuforia Engine SDK.

Hey, today I also think about that solution. First I have discussed about the solution with Image Target, which later on seem to be unsuitable because image when i move to the right of the object (house), then the whole AR content will be disappeared because the target will not be detected by the hololens camera.

Furthermore I just realize that I definitely have some 3D Model ob the object, which need to be fill with AR content, so I think I could make use of that.

So from your advice and from my understanding so far I will have to follow these step to do my work

1. Follow this instruction to create a model target from here (https://library.vuforia.com/content/vuforia-library/en/articles/Solution/how-to-create-model-target.html)

2. Import the newly created model target to unity like this instruction (https://library.vuforia.com/articles/Solution/introduction-model-targets-unity.html)

3. Start tracking the model in realview (the house in this case). In case the house got detected. User will see the model of that house via the hololens camera similar to the gif of this link (https://library.vuforia.com/articles/Solution/introduction-model-targets-unity.html)

An example from Vuforia PTC look like this (just under android):

https://www.youtube.com/watch?v=2ooSQmMrg4g

As I can see in the youtube video, when the camera detected the real model, the real object has been replaced by its 3D Model  (AR like)  and with the functions the model target generator provided, I could expand the 3D Model to use for inspect.

In my case of course I dont want to inspect the 3D Model. What I wish to have is:

After define the model target with external Application (Model Target Generator) and import this to Unity and deploy the Application for Hololens, I can start the App in hololens.

As soon as the real object(example house, table ...) got detected, instead of showing the 3D Model of the object (like in the video), the app should only enable the AR Content (Information of that object). So the user still only see the real object (object will not changed to 3D Model)  and the AR Content (information) as additional.

 

What I thinking of:

1. The Model Target Generator is a completely independent application. 

2. Similar like the question from the beginning. I have to do the import  of the to database manually (because of static device) so i can track the object later on.

What I really hope and wish to have an answer:

1. Is it possible to have the whole process from creating the target model until import to the database working in one app (special for Hololens). Is that possible with cloud targets like you mentioned above ?

Otherwise I have to follow the 3 steps and that mean I will have to use a PC or an android smartphone to construct the model target first and import that to unity so after that I can deploy to Hololens.

 

To sum up my whole unclear question: I would like to implement similar/same function like this example from vuforia ptc:

https://www.youtube.com/watch?v=y70yStPCBHA

except the target I aim to is the application should run on hololens and maybe if possible instead of showing 3D Model of the bike, only informations about the bike should appear.

 

 

Thanks for listening

 

 

Detect Image Target via Hololens Camera

March 7, 2019 - 9:03am #2

Hello,

quangbui wrote:

I would like to create an AR App for hololens, which can show informations about some specific thing (AR like) in the realworld, which is a 3D Object (eg. a big house) when i look at it.

for example when i use the hololens to view a house. i would like to see the information about that house like size, building time etc...

So like from the basic tutorial example from vuforia (https://assetstore.unity.com/packages/templates/packs/vuforia-core-samples-99026), I would say:

Firstly I need to define a 3d model target of that house via model target generator or maybe Object Scanner???. then use it as 3D Model Target in Unity and when the  hololens camera detected the model target, i can start to show the specific information (eg. size, building time, etc...).

You don't necessarily need to use Vuforia 3D target tracking technology to recognize a house. For example, you could take a picture of the facade of the home as a 2D target and use that for recognition. Note that this could be challenging for our tech due to changing environmental conditions (time-of-day lighting, shadows, seasonal foliage), and that similar home models could introduce false positives. However, if you combined the recognition with some location tech, such as GPS, then this could be resolved.

If you had a 3D model of the home you could certainly use that for Model Targets, and is likely a good use case. I'm assuming the challenge here is to get a 3D model that can be imported into the MTG, and is accurate enough to be recognized by the Vuforia Engine SDK.

I would eliminate the Object Scanner due to size limitations - it would be very difficult to scan a house. See this article: https://library.vuforia.com/content/vuforia-library/en/articles/Training/Vuforia-Object-Scanner-Users-Guide.html

quangbui wrote:

For me the house are like the Image Target which is similar to the cards from vuforia samples (but in 3D so it has to be model target) and the information which need to be show  will be in ar view.

How do I suppose to do that ? From my understanding I have to save the House as Target (which one to fulfill the job, i still cant figure out) in vufodia target manager.

Agree that the approach to use the house's facade as a 2D target is the quickest approach. With HoloLens, the significant challenge you'll have is with the device's spacial mapping range. This will have to be balanced with the 2D targets view to insure the house can be recognized by HoloLens, and the environment mapped. In other words, if the 2D image of your house is the entire front, then you'll need to be far enough away from the home for the HoloLens' webcam to see the entire home - this is likely too far away for spacial mapping. Thus, the home's 2D target view may need to be closer (only showing front door?) for HoloLens tracking.

quangbui wrote:

Do i have to define the Target manually with vuforia target manager everytime when i want to add a new target to my application, which need to be detected or can I do that dynamically via programming. 

Vuforia Image Targets are contained in static device databases that must be authored on the Target Manager. Cloud Targets can be accessed via RESTful APIs for creation, modification and deletion:

https://library.vuforia.com/content/vuforia-library/en/articles/Training/Cloud-Recognition-Guide.html

https://library.vuforia.com/articles/Solution/How-To-Use-the-Vuforia-Web-Services-API.html

Thanks,

Vuforia Engine Support

Log in or register to post comments