[PR] Creating basic image classifier with Elixir

[LuchoTurtle]

closes #1

This creates a simple image classifier with Bumblebee and performs image captioning tasks that return a descriptive text of the image (not to be confused with image classification, which yields a weighted list of predictions).

[LuchoTurtle]

I'm having immense trouble previewing the image and copying it to the DOM before consuming it. In order to actually use the image to classify it, I need to consume it. However, this effectively removes the image entry from the temporary folder and I lose access to it to show to the person.

I've tried using the image_live_preview function but it only works if the entry is not consumed, which is not possible in our scenario.

I'm trying to circumvent this by using an additional <input>, copying the file contents and displaying it in the DOM. But this, for some odd reason, makes it so that handle_progress/3 is not called, which in turn doesn't consume the entry.

allow_upload(:image_list,
       accept: ~w(image/*),
       auto_upload: true,
       progress: &handle_progress/3,
       max_entries: 1,
       chunk_size: 4_000,
       max_file_size: 64_000
     )}

I've tried many ways of doing this but I can't seem to figure it out. I could copy the file path to an upload directory, but that would mean that I'd need to manage all of the uploaded images, their correct deletion (even if the person crashes the LiveView), which is something I really don't want to "waste" time on.

[nelsonic]

@LuchoTurtle maybe a quick sketch (paper notebook flow diagram) of what you're trying to do will help others on the team understand where you're stuck. 💭

[LuchoTurtle]

I've added virtually all of the documentation and the application is now functional. I'm just having some trouble with having the tests covering 100% of the application. Because I'm using Task.async, I could technically add a sleep method for it to wait for the model to predict a given image. However, I feel like that's a "hack"/dirty copout.

I've been researching on how to get this to work, namely form:

• https://elixirforum.com/t/cant-use-task-async-within-liveview/32940/2
• https://medium.com/elixir-learnings/testing-a-phoenix-liveview-that-does-an-async-operation-after-mount-b8ec27e6c167

But haven't had good results on this.

I'm trying to implement a supervisor to do this (from https://elixirforum.com/t/best-way-of-doing-an-integration-test-that-spans-a-task-start/39429/2) but I'm struggling to get it to work.

Other than that, the app works (although it does not show the image after submitting, which is something I wanted to get to work as well, but it's not "essential")

[codecov[bot]]

Welcome to Codecov :tada:

Once merged to your default branch, Codecov will compare your coverage reports and display the results in this comment.

Thanks for integrating Codecov - We've got you covered :open_umbrella:

[LuchoTurtle]

This should be mergeable 👌

[nelsonic]

@LuchoTurtle demo on Standup this morning was superb! 😍 PR is still assigned to you. Happy to review it later today. 👌 Meanwhile there are a couple of tasks you can crack on with: #3 & #4 😉

[LuchoTurtle]

I've reassigned it to myself because I want to make some additional changes. I'll hand it over to you after those.

[nelsonic]

[LuchoTurtle]

Reassigning back to you 👌

[nelsonic]

Gonna pick this up (run + review) as soon as I get a decent chunk of alone time.... 🤞 ⏳