Machine Vision | Deep Neural Network Architectures
Toronto Transit – Fare Evasion
According to TTC’s Audit, Risk and Compliance report (2019), fare evasion on streetcars costs $25M annually.
Sheyld AI completed a project using deep learning computer vision techniques to count the number of people present in a video feed. As a result of this solution, the TTC would be able to compare this estimate to the fare collected for the day, prioritize resources and control costs.
Deep Learning Architectures
Object detection is a task that focuses on locating and classifying objects in an image. Unlike other traditional machine learning techniques, deep learning is better suited for this task, as it is able to identify complex patterns in this type of unstructured data.
We compared three object identification algorithms, including:
Finally, we determined the benefits and trade-offs of each approach by using performance metrics such as mean-average-precision (mAP).
As a result of the analysis, we recommend the use of mask R-CNN with transfer learning for this application. We were able to achieve a mAP of 77%.
The reason we want to use Mask R-CNN over YOLO is that the former has the added benefit that could be used for density mapping, a method used for counting or estimating the number of people on the streetcar.
New revenue opportunities!
The implementation of an AI model will be the first step in transforming the TTC’s operational architecture into a data-centric organization. Additional revenue opportunities can be achieved by improving:
Identification of fare evaders
Automation of the ticketing process to reduce long processing times and increase revenue
Implementing both AI and automation will allow the TTC to reach 27% increase in profit in the long term.
We offer the leading edge end-to-end AI solutions, enabling every employee, customer, and citizen with sophisticated AI technology and easy-to-use AI applications. With our solutions, our partners can become leaders in their respective domains by becoming more innovative in serving their customers, collaborating among all key stakeholders in extracting business value and reducing costs.