Final Project for Implementing ANNs with TensorFlow
WS 21/22, Osnabrueck University
Dennis Hesenkamp, Lennart Zastrow, Madhuri Ramesh
This repository contains all files associated with our final project for the abovementioned course. Idea and data come from Zindi's Turtle Recall: Conversation Challenge.
To continue, make sure that you have a local distribution of Conda installed (e.g. Anaconda, miniconda, miniforge). The code in this repo has been written and tested on an Apple Silicon device. Create a local environment with the environment.yml file to be able to execute code from the turtleRecall.ipynb notebook.
data: Contains dataframes as .csv files with image IDs and associated turtle IDsdocumentation: Written documentation of the project, including LaTeX source filesfunctions: Helper functions used in the main notebookmodels: Code to create the models we usedenvironment.yml: Environment in which this project was builtturtleRecall.ipynb: Main notebook with pipeline for the entire projectturtleRecallColab.ipynb: Main notebook adapted for use with Google Colab. Works as standalone, all helper functions, models, and utility variables are includedutils.py: Utilities
The overall pipeline can be executed following the turtleRecall.ipynb notebook. The notebook contains useful notes and explanations and guides through the entire process of data acquisition, preprocessing, data augmentation, model creation, and training. It can be seen as a helpful complement to the documentation (documentation/documentation.pdf), which describes many of the steps in greater detail and with scientific background.
Use the turtleRecallColab.ipynb notebook for execution in Colab as opposed to local execution. The image files are stored online in a GBucket, which makes working with them in Colab easy, no local download required.