Create a diagram showing the relationships of all the objects

[hamlinliu17]

Simple google drawing that is the UML diagram of all the objects. Will update my progress and post questions on this thread.

[hamlinliu17]

Link to the drawing here

[hamlinliu17]

@vc1492a Here's what I have from looking through the modeling.ipynb and dl_timing.ipynb. Let me know if I am missing anything important

[vc1492a]

@hamlinliu17 thanks for putting this together! Sorry for the delay in getting to take a look. This looks great! Just a few small comments:

• For inference, I don't see storing the generated images to disk as the default behavior in real-time. In a real-time scenario, we'll simply want to generate the image and then process the image directly without saving and reading from disk (causing unneccessary I/O). Maybe we could implement a flag that by default is false, indicating that the images are not saved to disk.
• Predict images in inference would load the images and use the model stored in memory (loaded once) to make the prediction. I could see inference being used in a simple microservice / API down the road.

The rest looks aligned with what I was thinking. This is good documentation, let's make sure to include it in the repository (or link to an editable copy). 👍

[hamlinliu17]

Reopening this issue since I will need to use this to write down and record some ideas as well as post images of my progress. Attached is the UML diagram of the model.py (link to drawing url)

[hamlinliu17]

@vc1492a
adding a link to where I am currently visualizing everything here. It is still a work in progress though

[hamlinliu17]

Here is a nice look at it.

[hamlinliu17]

@vc1492a From how I was looking at it, I think we can encapsulate a little more of the out of sample experiment maybe have a function to encapsulate the image prediction part and maybe move the read_data somewhere else? Also would like to visualize how we are going to build the pipeline through a diagram when we dig deeper.