UAV Tracking for Smart Cities

Darshan Lakhankiya
Sneha Seenuvasavarathan
Vedant Vohra

- assets/ # Contains all the assets
- data/ # Contains the data files
    -> /sample_data/ # Contains a small subset of the dataset for testing the model ("grading script")
- models/ # Contains the trained models
- notebooks/ # Contains the main project notebook(s) and the grading notebook

NOTE: The dataset (P-DESTRE) is too large (69.4GB) to include in the zip file, so it has to be downloaded separately. Please refer to Step 1 in the 'How to Run' section below for instructions.
A small subset of the dataset (data/sample_data) has been included in the zip file and can be used to run the grading script, without having to download the entire dataset

This directory contains all the source code in the form of Jupyter notebooks.

1. reid_train.ipynb - This is the notebook used for training the pedestrian re-identification model
2. reid_pipeline.ipynb - This is the notebook containing the end-to-end pipeline for pedestrian detection and re-identification
3. test.ipynb - This is the "grading script". It is used to test the model on a small subset of the dataset.

1. Execute data/download.sh to download the dataset.
   Alternatively, you can download the 2 files from the following page and extract them to the data/ directory:
   http://p-destre.di.ubi.pt/download.html#:~:text=CVPRW.2018.00281%2C%202018.-,Download,-This%20dataset%20is
   Make sure to extract the contents of the rois.zip file into the data/P-DESTRE directory.

2. If docker-compose is installed, run:

   docker-compose up

   This is the recommended way to start the container.

   If docker-compose is not installed, then execute the following command:

   sudo ./run.sh

3. Once the container is up and running, you will see a URL displayed in the terminal. Click on it to open a jupyter session in the browser.

4. Open the notebooks in the notebooks/ directory in Jupyter.