UAV Billboards - ML training

The instructions below are for training the model and evaluating it on the test set.

Requirements

python3 - for running the scripts
```
sudo apt install python3
```
ffmpeg - for video conversion
```
sudo apt install ffmpeg
```
exiftool - for metadata extraction
```
sudo apt install libimage-exiftool-perl
```

Install the required python packages using pip:

pip install -r requirements.txt

Download dataset using download.py script:
```
python3 dataset/download.py
```
Convert dataset from COCO to YOLO format:

Note: You have to select do you want to generate bounding boxes (b) or segmentation masks (s). We used segmentation masks in our experiments.
```
python3 dataset/convert.py
```
Split dataset into train and test sets:

Note: The script below provides many options. Note that we use them in the following way: y (delete the existing data), s (segmentation masks), y (crop the images), gps (split the images based on geotags), n (no merge billboard classes), n (no exclude road-sign class).
```
python3 dataset/prepare.py
```

YOLOv7 and YOLOv8 models are training using the scripts and configs from the yolov7 and yolov8 directories.