Health and Safety are serious considerations for the industry given 260 deaths over the past year, and 120,000 serious injuries. People can be complacent, in a rush, or have gaps in their H&S education. We’ve built an AI model that is trained to learn what dangerous behaviour looks like, and overtime can use unsupervised learning to pick up H&S risks automatically from a variety of video camera feeds.
Saving lives & Preventing injuries
We began with a simple showcase prototype that addressed the most common area of injuries and potential deaths: loading steel on flatbed trucks. We took 44GB of video footage and created an AI model that informed supervisors when workers were not wearing the appropriate safety gear, so they can better target their education programmes without being invasive.
A Datacomp Story
We got this process started by including the truck monitoring safety solution as a challenge in Datacom’s annual hackathon event (Datacomp) in 2017.
During the hackathon event, the Datacom team prototyped and trained hard hat real-time detection. The core aim of the initial problem was to unlock insights from previously untapped sources of data.
We analysed 44GB of video footage to get an AI model trained to a very high degree of accuracy
Built an alerting system which notifies managers of risky situations
There is significant opportunity for this solution to have a broader market impact