Object Detection Python* Demo - code comments

[anujonthemove]

Hello,

This may or may not be the right think to ask but the earlier version (Object Detection SSD Python* Demo) of the code was quite straightfoward and self explanatory. This version no doubt is much more organized but it is not so easy to understand therefore I'd like to know if there could be a code flow diagram or at least steps as to what is happening around. Also, for things like utilization_monitors, the given description does not provide a very clear explanation of how it can be used.
Please do point me to the right resouces.
Thanks

[vladimir-dudnik]

@anujonthemove thanks for feedback, this is right thing to ask and point if something is not clear in demo code. Actually, our intend was to unify and further simplify OMZ demos with having class hierarchy to express that there are some basic steps which are required for any model to run inference:
-process cmd line params
-set data source
-preprocess data
-run inference
-postprocess the results
-output results

I would not say we've made a final implementation, but at least we are going in this direction.

@Wovchena @eizamaliev @fzhar FYI

[eizamaliev]

Hi, @anujonthemove

Thanks for feedback! Could you tell us more about your vision of code explanation? What would you prefer, the comments, the description how code work in the readme.md file, or something else? Which parts of code are the most complicated?

[anujonthemove]

Hello @eizamaliev

Ideally code comments are enough in most cases but here I wanted to understand the code flow as you suggested by giving a clear description of how the code works in the readme file.

Following are the things which I could not comprehend easily:

1. The use of --utilization_monitors and at least how to use it?
2. In the beginning of the first while loop, what's the need to check for callback_exceptions?
3. Then the object presenter which calls the drawGraphs. Since the monitor module is completely in C++ so there should be an explanation of how we are making use of it and a brief overview, if possible.

I am trying to understand the pipeline piece by piece as my end objective is to make an API out of this object detection pipeline.

[anujonthemove]

@anujonthemove thanks for feedback, this is right thing to ask and point if something is not clear in demo code. Actually, our intend was to unify and further simplify OMZ demos with having class hierarchy to express that there are some basic steps which are required for any model to run inference:
-process cmd line params
-set data source
-preprocess data
-run inference
-postprocess the results
-output results

I would not say we've made a final implementation, but at least we are going in this direction.

@Wovchena @eizamaliev @fzhar FYI

Thanks @vladimir-dudnik thanks for outlining the basic steps. I am happy to learn that more development is on it's way!