Product Development
The Goodyear Tire & Rubber Company, a global leader in tire manufacturing, sought to explore alternative methods of tire management for fleet operators.
Traditionally, handheld readers and ground-based sensors depend on technology that uses laser illumination or similar methods to read and report data such as tread depth, sidewall condition, and more. While effective, these technologies involve expensive equipment that is not easily accessible to all fleet operators.
Goodyear wanted to test if smartphones could deliver the same value by leveraging image capture and computer vision technology. This new approach would provide a more accessible, user-friendly solution for fleets without relying on expensive or specialized equipment.
The company sought a trusted software development partner to run a proof of concept (PoC) experiment. The goal was to evaluate the ability of in-market smartphones to meet three goals:
1. Read and analyze tread depth using in-market smartphones.
2. Report on the condition of tire tread, categorizing it as “Good,” “Getting Worn,” or “Change Now.”
3. Validate the accuracy of smartphone-based readings against traditional methods.
Goodyear partnered with Taazaa to develop the PoC for this smartphone-based tire analysis application.
Taazaa rapidly built the PoC by developing a computer vision solution for tread depth measurement. The solution involved a series of well-designed experiments to assess if smartphones could reliably capture and interpret tire tread images.
Our team first gathered a diverse set of tire images representing various tread patterns, wear levels, and conditions. Each image was carefully labeled and annotated to assign tread depth values, ensuring high-quality data for training the machine learning model.
We employed advanced image processing techniques to preprocess the data (e.g., grayscale conversion, normalization) and extract relevant features such as edges and contours. The team then developed a machine learning model to predict tread depth, utilizing convolutional neural networks (CNNs) for superior image analysis performance.
We split the dataset into training, validation, and test sets to ensure accurate model evaluation. The model’s performance was measured using key metrics such as Mean Absolute Error (MAE) for regression tasks and accuracy for classification tasks (i.e., categorizing tread depth as “good,” “getting worn,” or “change now”).
We integrated the trained model into the smartphone image processing pipeline, allowing real-time prediction of tire tread depth. Continuous calibration with real-world measurements ensured the system’s accuracy and relevance to Goodyear’s needs.
By delivering a functional proof of concept within 12 weeks, Taazaa enabled Goodyear to assess the feasibility of smartphone-based tire tread analysis. The speed of delivery helped Goodyear make informed decisions quickly.
The project demonstrated that smartphones can effectively read and analyze tread depth, providing real-time feedback on tire conditions without requiring specialized hardware.
It also proved that machine learning models can achieve the desired level of accuracy in predicting tread conditions, enabling Goodyear to pursue further development in this space.
The successful PoC opened the door to the development of cost-effective and accessible solutions for fleet operators, reducing the need for expensive, specialized equipment.
The solution’s flexibility allows Goodyear to scale the technology for broader use in the future, integrating it into their existing product portfolio.
Through this collaboration, Taazaa helped Goodyear validate the potential of smartphone-based tire management systems. The project showcased Taazaa’s ability to develop innovative, AI-driven solutions tailored to client needs.
This proof of concept laid the groundwork for future technological advancements in Goodyear’s fleet tire management services.
.svg)
