Technical Debts

[ktl014]

List below things we need to refactor for the hab-ml codebase

[ktl014]

1. Standard logger for logging error/debug msgs
2. Get_image_dir un_hardcoding during model deployment
3. Joining csv files and jsons together after deployment
4. giving read/write access for all main directories
5. argument parser vs config file

[ktl014]

1. probabilities in csv files should only take top 5 probabilities

[zhouyuanyuan1999]

1. Add features to be able to query the right model_dir easier. Improve model_dir naming convention.

[ktl014]

1. add unique id to predictions.json file in deploy mode

[ktl014]

Update for number 5:

• Need to have different configurations when developing locally vs testing on the lab machine.

@igneel73 any comments on this?

[ktl014]

![Uploading image.png…]()

[ktl014]

Remove spc submodule from hab_ml

[Eighteensi]

I would like to replicate this model, but I do not have the image recognition data and labels. Do you still have the original data from back then? I would like to try out the model's performance for further research. Thank you