carlowiesse / raindropdetection Goto Github PK

SUMMARY:

This work proposes a solution to droplet classification using Gist features and an SVM classifier. This solution achieved perfect recall and precision scores on both training and evaluation sets. As a results, the submitted python file contains this method as my final solution to the task. Additionally, this work explores 2 other alternative approaches (deep learning and blob detection) for the sake of comparison. The implementation of these 2 alternatives can be found in the jupyter notebook files and their performance is analyzed together with the proposed approach in the attached report.

========================================================================== CONTENTS:

Code development was done using Python 3 on jupyter notebook files. The final submission consists of a'.py' file containing the source code together with a necessary '.pkl' file containing the trained classification model for the proposed approach. Additionally, there is a PDF report explaining the steps I took to arrive to my proposed approach, highlighting the decisions I made along the way.

========================================================================== FOLDERS:

models/ This folder contains 10 trained classification models, 5 intances of an SVM model (Proposed approach) and 5 instances of a deep learning model (alternative 1).

notebooks/ This folder contains all the jupyter notebooks I used during the development of this project. They include code with comments and short paragraphs about the steps I took and what I learnt from each taken step.

reference/ This folder has PDFs of the scientific papers that I referenced in some jupyter notebook files and in the final report.

report/ This folder holds the files I used with Latex to produce the PDF report. The final PDF report is included in this folder as well.

solution/ This folder is the package that contains my proposed solution to the classification task. It includes a '.pkl' file with the trained ML model and a '.py' file with the source python code ready to be run. (Make sure you have the required libraries installed)

tools/ This folder has python scripts that I used in the jupyter notebook files.