BUSINESS SITUATION
SGA was engaged to develop an automated system that can identify and estimate the location, size, and availability of parking spaces to assist drivers in easily locating parking spaces while driving. The solution was built to work efficiently on live camera streams while optimizing speed, accuracy, and other parameters.
SGA APPROACH
- We designed problem formulation, analysis, and determination of data requirements.
- Collaborated and partnered closely with the client’s teams for data definition and requirements, we also finalised class definitions needed for model training, data gathering, and ingestion.
- Our team conducted primary and secondary research, a literature review, and the design of experiments.
- We created a design thinking-based solution architecture and framework development to support the methodology for data arrangement, pre-processing, and data transformation into a model-ready format.
- Our development team trained deep learning (DL) algorithms for object detection, distance, and size estimation, to assess model performance.
- Once we implemented object tracking algorithms to optimize distance and size estimation of parking space to reduce redundancy, our team performed multiprocessing for quicker execution and results.
-
ENGAGEMENT
- Our team first created an initial proof-of-concept followed by a phase-wise delivery. This was a hybrid engagement with a weekly outcome plan aligned with the overall project objectives.
BENEFITS & OUTCOME
- A highly accurate system was developed to automatically identify the location, size, availability, and orientation of parking spaces, which was optimized for real-time detection.
- The solution is a part of the driver assist system with potential large-scale commercial adoption.
KEY TAKEAWAY
SGA helped the client enhance its driver assist system with additional automated driver assist features.
