Machine Learning CMT307 coursework project Group 12
You found this script inside the folder cmt307-g12.
For this script to run, you must include a data folder inside the same directory (cmt307-g12 folder).The data folder can be downloaded using any of these two links:
Full version (recommended) The link below contains all the source pictures to perform the analysis, but also the checkpoints from our own notebook runs. Original and transformed pictures will be loaded from these files instead of being built from the notebook, saving considerable time. 6.7 GB.
Full version - Google drive shared folder
Light version (minimum requirements) The link below contains only the ppm pictures from the train and test datasets, and the labels from the test set. The pictures will be decoded and transformed in your own machine/Colab session. 563 MB.
Light version - OneDrive shared folder
If for whatever reason the links above don't work, you could also find the data folder in the github repository (light version).
Contents of cmt307-g12 after you include the data folder:
| Item | Description | Required to run |
|---|---|---|
| main.ipynb | Present python script | Yes |
| data | Folder containing data sources and pictures | Yes |
| report | Folder to save pyplot pictures to be used in the report | No |
| README.md | README file with instructions | No |
| requirements.txt | txt file listing required python modules/packages to run the script main.ipynb | No |
This script is capable of detecting whether you are working in your local machine or in Google Colab, and will adjust the data directory addresses accordingly (section 0.2).
Download the data folder from the above link and paste it inside cmt307-g12.
Then the script can be run using Jupyer Notebooks from whatever location.
Download the data folder from the above link and paste it inside cmt307-g12.
For this script to run in Google Colab, the folder cmt307-g12 must be placed in the ROOT of your Google Drive:
%cd /content/gdrive/MyDrive
This project has been uploaded to github as a repository. You can clone this repository in your local machine or in your Google Drive:
https://github.com/jm20389/cmt307-g12
If you run this repository in Google Colab, the repository folder must be cloned in the ROOT of your Google Drive:
- In Google Colab, mount your drive first:
from google.colab import drive
drive.mount('/content/gdrive')
- Navigate through the Google Drive root:
cd /content/gdrive/MyDrive
!ls
- Clone the repository using Git:
!git clone https://[email protected]/jm20389/cmt307-g12.git
If you run into RAM issues...
Due to the magnitude of this project, and depending on your machine or cloud session setup, you may run out of RAM memory. During section 0.2.2 , recovery functions are defined, to retrieve picture data from the checkpoint files (stored in the subfolder data/numpy).
In the unlikely event that your session crashes, and provided you have all the checkpoint files available (either downloaded from the links provided or created from running sections 1 and 2), you can resume the analysis from Section 3 as follows:
From a fresh session, run sections 0.1, 0.2.1 and 0.2.2, then open a new cell anywhere and run the function RecoverEverything() . All required items to carry on from Section 3 will be loaded in the kernel to continue the analysis.