Disclaimer: Aya Data holds to the confidentiality agreements made with our clients. We shall not disclose any confidential information regarding clients and/or projects, pursuant to the individual agreements made. Consequently, client and project names and any identifiable information may be kept anonymous in our Case Studies.
The global electric scooter market size is expected to reach USD 34.7 billion by 2028, expanding at a CAGR of 7.6%. Needless to say, this is changing the landscape of transportation, especially in urban areas. The growth will impact public transportation, city infrastructure, and critically, how self-driving cars should be trained to ensure the safety of the people around them.
The rapid surge of rideshare e-scooters on city streets has so far been unmatched by advances in regulation, and city councils are still trying to understand what the best measures to utilize the technology are, whilst keeping people safe.
As this market continues to grow it is critical for AI companies to keep up with the capabilities and prevalence of short-distance vehicles, the laws and regulations, the common human behaviors, and how to ensure AVs can continue to operate safely in our cities.
The Client wanted to cover off an unaddressed edge case in their AV training workflow – the burgeoning existence of e-scooters on European roads. Open-source datasets could not provide sufficient instances of this edge case or were labelled inconsistently.
The challenge was to create a large and purpose-built dataset that would enable AV models to detect, classify, and ultimately predict the movements of e-scooters.
Aya Data’s AV labeling experts labeled 10,000 images of e-scooters in different contexts, positions, and situations. The e-scooters were bounded and tagged with classes denominating the type and associated properties.
The images labeled by Aya Data were used to train the Client’s computer vision model. The end result was that the CV model could detect e-scooters in real time with 95% accuracy.