SCIA-Premium / DNN

This repository contains the project of DNN class at EPITA. The goal of the project was to benchmark multiple type of Visual Object Trackers on multiple dataset using different metrics.

DNN

AUTHORS

Alexandre Lemonnier <[e-mail]>
Alexandre Rulleau <[e-mail]>
Baptiste Bourdet <[e-mail]>
Eliott Bouhana <[e-mail]>
Marius Dubosc <[e-mail]>
Philippe Bernet <[e-mail]>
Sarah Gutierez <[e-mail]>
Victor Simonin <[e-mail]>

---

The goal of the project is to analyse this . Then you will have to select at least 6 trackers and compare them qualitatively and quantitatively on different datasets.

Installation

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Description

This github contains multiple notebook to execute in a specific order.

---

The following notebooks creates .py files that are used by the other notebooks.

• dataset : Download the dataset and extract it with a specific structure
• model : Load the models with their default values from different github
• scoring : Define function to score the models predictions

---

The following notebook will create folder containing the results of the benchmark.

• benchmark : Run the benchmark for each model on the dataset using the scoring functions

---

The following notebooks will create graphs and tables to analyse the results of the benchmark.

• benchmark_stats : Compute the statistics of the benchmark using the files generated by the previous notebook