Developing a Proof-of-Concept

A proof of concept is an important tool for demonstrating the possibilities and value of Pacenote.ai. The proof of concept will also be the foundations for the final application as we can build upon the work already done. When deciding to build a proof of concept, it’s important to consider the scope of the concept. For example, if too much complexity and functionality is designed into the proof of concept, it’s going to take a long time to develop. However, too little functionality is not going to accurately demonstrate the aim of the application. Generally, simplicity is key so the proof of concept needs to be as simple as possible while still showcasing it’s ultimate aim.

For Pacenote.ai, the proof of concept was designed to tackle the core functionality of the anticipated use case. In simplistic terms, the POC needed to be able to process a video of a vehicle driving a long a road section and be able to detect turns made in the video and the severity of each turn. This scope was deemed to roughly encompass how rally crews might use the application: take a video of a recce on their phone from inside the car and give it to Pacenote.ai in order for it to output a series of pace notes which will include direction of turns and the severity.

Just this simple scope also calls into question some other important functionality that we decided to include in the proof of concept. For example, rally crews might use a variety of camera equipment to record the recce so we needed to include functionality to calibrate the input video in order to accept video from any type of camera.

Additionally, rally crews have their own unique ways of categorising the severity of turns on a rally stage. For example, one crew’s “five right” is likely to be very different to another crew’s interpretation. This meant the proof of concept needed to include a way for the user to define the categories of turns.

The proof of concept masterfully handles all of these requirements. During comprehensive testing, it’s shown to be more accurate than anticipated and easier to use as well. The POC utilises Simultaneous Localisation and Mapping (SLAM) in order to read the road within the user’s video. SLAM is a technology within the broader category of computer vision that is often used within robotics and self-driving cars. SLAM allows the application to understand it’s location within a video (or single frame of a video) and also create a map of that location. For the pacenote.ai proof of concept, it’s able to detect directional changes between frames of a video. It can also detect how severe the directional change is and the duration of the change (e.g long corners vs short corners).

The proof of concept is viewable now and we are keen to show it to anyone who’s interested. We are comprehensively testing the proof of concept and would like to show it to interested rally crews in order to gauge their interested and gather their feedback. Please get in touch if you’d like to know more.