RVSS Tools Scripts Documentation

Collection of Python scripts for things relating to the RVSS_Need4Speed challenge.

Scripts Overview

vid_stitch.py

Creates videos from image sequences with angle overlays.

• Processes sequences of images named in format 000001_angle.jpg
• Adds angle information as text overlay
• Outputs MP4 video with configurable FPS and duration
• Usage: python vid_stitch.py --input_folder /path/to/images --output video.mp4

dataset_clean.py

Dataset visualization and cleaning tool.

• GUI interface for reviewing image datasets
• Shows angle information from filenames
• Allows deletion of unwanted images
• Useful for cleaning training datasets

test_model.py

Testing script for trained PyTorch models.

• Tests model predictions against ground truth angles
• Processes test images named as 000033_0.20.jpg format
• Configurable model path and test folder
• Usage: python test_model.py --model_path model.pth --folder_path /test/images

bot_stop.py

Connects to the robot and sends a stop command if you get a crash when collecting data or deploying a model and the robot just keeps driving

• Usage: python bot_stop.py --ip 192.168.1.xxx

get_image.py

Requests a single image from the camera onboard the PenguinPi and saves it as image_x.jpg depending on how many images you have grabbed before. Useful for grabbing images of the stop sign for fine-tuning thresholding or some other method.

• Usage: python get_image.py --ip 192.168.1.xxx

check_battery.py

Request a battery reading from the PenguinPi and prints the result as a calculation of the typical voltages for 18650 battery cells.

• Usage: python check_battery.py --ip 192.168.1.xxx

dataset_csv.py

Creates a CSV file of images and labels for a given folder. Useful for some existing tools for analysing image datasets.

Installation

If you've already setup your RVSS mamba/conda environment, you can simply clone this repo and then run these scripts from the RVSS_Tools directory.

git clone https://github.com/TedVanderfeen/RVSS_Tools.git
cd RVSS_Tools