https://github.com/monamoxie/python-machine-learning-image-manipulator

A python based machine learning app for image manipulations.
https://github.com/monamoxie/python-machine-learning-image-manipulator

Last synced: about 1 month ago
JSON representation

A python based machine learning app for image manipulations.

• Host: GitHub
• URL: https://github.com/monamoxie/python-machine-learning-image-manipulator
• Owner: Monamoxie
• Created: 2024-02-05T19:14:05.000Z (12 months ago)
• Default Branch: master
• Last Pushed: 2024-07-16T18:42:50.000Z (6 months ago)
• Last Synced: 2024-07-16T22:53:17.103Z (6 months ago)
• Language: Python
• Homepage:
• Size: 42 KB
• Stars: 0
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# IMAGE RUNNER

Work in progress (WIP) 


An image manipulator built on

 - Python

 - Fast AI

 - OpenCV

 - Matplotlib

 - Django (for UI & Data Management)

## TRAINING THE MODEL LOCALLY

- The training of each model will be handled through the CLI. You can find this in the `management/commands` directory of each app.

-  For instance: running `python manage.py cat_dog_trainer` will run pre-train the model and save it to the mlmodels directory. 

- This is to be called once, usually during the first setup. 

- Subsequent requests from the web ui will be routed to the trained model via a service layer. 

## USING CLOUD BASED GPUS

- Due to how resource intensive these operations can be, and depending on your hardware, it is always advisable to use cloud platforms like Google Colab, Sage Maker etc. 

- Each CLI class has a private method in it called `__execute()`

- You should copy the content of that method, as well as all necessary imports at the top of the class, and run them in your cloud platform. 

- You should endeavour to change the path of the exported path. For instance, instead of setting your path as this:  `model_path = os.path.join("image", "mldmodels", "cat_dog_model.pkl")`, you can choose to set it as this: `model_path = '/content/model.pkl'` 

- If using Google colab, you could choose to connect to your drive and mount the file to it, then export the trained model when completed. 

- All trained models should be placed in the `mlmodels` directory of the respective apps.  

 # To use the GUI capabilities of OpenCV with this Docker container:

1. Install XQuartz (Download Link)

2. Click on `Preferences` from the Menu bar and toggle on `Allow connections from network clients`

3. Vist the root of the cloned project and run `xauth list`

4. Copy the output from the above step

5. Run `echo "" | sed -e 's/^..../ffff/' > .docker.xauth`

6. This should create a file  `.docker.xauth`. This file is already mapped and set within the python service

7. Run `xhost +`