This repository contains the solution for the VRRT bachelor thesis.
This repository is structured as follow :
data-- This folder contains all the data necessary to do the project. For now, the data is not open source, this can change in the futuremodels-- This folder contains all the ultralytics YOLO models. Some downloaded as is from their website, some fine-tuned for the taskresults-- This folder contains all the results for the project. For now, the data is not open source, this can change in the futuresource-- This folder contains all the sub-projects. Each sub-project contains a README with more detailstools-- This folder contains the scripts created during this projecttrackers-- This folder contains the different trackers configuration used in pair with the YOLO model
The data, models, results and trackers are linked in the subfolder to allow sharing of folders between the differents subfolders
For the installation of the sub-projects, you can refeir to their README. But they all use the UV package manager
The bib_detection sub folder contains all the code for fine-tuning a YOLOv11 model, completing and inspecting a YOLOv11 dataset.
The stereo sub folder contains all the code for calibrating a stereo setup, rectifying images, estimating depth using diverse algorithms.
The record sub folder contains all the code for recording video using a single camera or a stereo setup. This include a 'best-effort' syncronization between two cameras for a makeshift stereo setup.
This is the code used to gather most of the data used in this project.
The project sub folder is the main project folder. This contains all the final pipeline, using the byproduct of the others subfolders (the data from ./record, the fine-tuned model from ./bib_detection)