Drawing only needs to obtain the attitude matrix after recognition and tracking.

Drawing, you can use Unity3D, or you can directly use OpenGLES for model rendering and model animation analysis.

Of course, the advantage of Unity3D is that it can simplify a lot of learning costs, and it takes a lot of effort to use OpenGLES. Moreover, there are not many open source codes available for rendering and animation, and many of them have to be written by themselves, which is not recommended by novices.

So the advantage is quick solution.

Applications developed based on Qualcomm sdk will have watermarks. If you want to remove the watermark, or if you want to use the cloud to identify the version, you have to pay.

Based on OpenCV.

There is no ready-made open source code for tag-free identification and tracking, and most of the open source codes that can be found are tag-based, which is very inefficient.

To do AR identification and tracking, you need to develop it yourself.

Moreover, this part of the learning cost and development cost are relatively high.

It is difficult to achieve the efficiency/accuracy/robustness of Qualcomm.

Although many papers can be found in this part.

However, various problems will be found in the implementation.

The advantage is that self-study ability is improved.

But it is difficult to reach the level of commercialization.