Keep Our Roads Clean
Helping municipal services run more efficiently by auto detecting litter bins awaiting collection, debris to clean and road cracks needing action.
As a give back to the communities we live in, Sensor AI built a rapid prototype for local municipal services to better maintain the
community envrionment while reducing the burden on taxpayers.
Business Overview
Our local councils appoint rangers to regularly patrol the streets looking for random road litter, illegal dumping, bins for collection and so forth.
Issues such as damaged pavement, broken billboards, fallen signs and fainted road markings can take prolonged waiting time, depending on how visible they are.
Patrol relies heavily on human eye observation and it can be a miss or hit.
Sensor AI has built a custom Machine Learning solution to help detect events such as open litter bins, illegal dumping, debris and damages requiring attention.
Feeding from the service van mounted camera recordings, the solution reports back detected anomalies and events for action the next day, significantly reducing maual workload and service turnaround time.
Our solution is developed iteratively through the practice of MLOps, releasing new features in small sprints and in agile fashion.
Afterall, happy community happy life !
Below are some examples of what our solution was able to detect from council's onboard cameras:
Road cracks
Illegal Dumping
Debris
Agile Development
Splitting training video clips provided by the councils into frames for annotating the events of interest, we managed to label over 2000 image frames per day using the labelling tool provided on Azure ML.
ML Models are developped in small batches, publishing one annotated event per training itertation via continuous integration and continuous release.
Visualize AI output
Every week council vans upload their patrol camera recordings. As part of the nightly process routine the solution detects events and aggregates results into a visual dashboard for attention and action the next day.
MLOps Architecture
Our AI delivery framework employs MLOps practices across all four stages of ML development lifecylcle. Combining the modern ML platform and CI/CD tools to automate training, testing and release processes allows our skilled data scientists to focus on what they do best, accelerating time-to-market.
