Smartvid.io’s construction-tuned AI improves safety and risk analytics
AI may seem like a magic, super advanced wonder of coding, and, to be honest, it is. But it’s not quite as magical as the general public may think. We don’t simply compile lines of code and have HAL 9000 pop out of the other end. Like any modern AI, our “Vinnie” requires a great deal of careful training to be able to examine an image and recognize people without PPE, cluttered walkways, work at height, and the other potential safety hazards.
So we’re happy to announce that Vinnie can now “see” a whole set of new things. Even better, these aren’t just new hazards. We trained Vinnie on tens of millions of images to give Vinnie the ability to recognize context.
For example, if Vinnie notices workers in a photo who aren’t wearing PPE, it makes a lot of difference whether these workers are participating in a demolition or giving an interior one last coat of paint before it’s turned over to a tenant. That said, while a lack of PPE at the finishing phase of a project may be no big deal, standing water would be an enormous problem.
Vinnie detects worker at height
(Photo courtesy of 2018 ENR Photo Challenge)
Here are the new construction activities that Vinnie detects automatically for every image:
- Concrete pour
- Trench work
- MEP (mechanical, electrical, plumbing)
- Structural steel
- Finish work
Finally, we have more tags related to personnel:
- Worker in cab: These images show someone driving a vehicle or equipment.
- Worker at height: Here, someone is in a lift, on a ladder, or near an exposed edge.
- Workers in groups: This is a way of identifying if there are groups of workers who may be standing too close to each other and not practicing “social distancing” in our new era of Coronavirus (available in early April).
Get 60 days of free access to the "People in Group" tag and other Safety Monitoring module features.
So, you may be asking, what’s the benefit to you?
First, you’ll get a lot more signal and a lot less noise in your analytics. It’ll be simple to see whether a high number of people without PPE is a serious safety risk or just a crew inspecting a finished room.
The new tags also provide opportunities for more efficient workflows. First, Smartvid.io can automatically recommend risk ratings based on the hazard detected in combination with contextual tags. Additionally, if there is demolition going on and you want to dive more deeply into the photos, you can find any demolition image with a quick search, click on it, and create an observation. You can also establish rules within these contexts. If your firm has a 100% compliance policy with high-vis gear during excavation, you can set that rule within Smartvid.io.
You can also get benchmarks by type of activity. Want to know what your PPE compliance percentage is during concrete pours? It’s simple to generate a report on that very stat, which can help inform the conversations you have with your safety manager prior to the next concrete pour.
Vinnie is already smart. Unlike human beings who would tire and lose focus after identifying hazards in 1,000 images, Vinnie never falters in its ability to find them quickly and accurately. And with these new tags, the AI can now internalize some of the context that people grasp without barely thinking — and it can do so much faster and with greater precision. The better Vinnie gets, the more information we’ll all have to identify risk and stop incidents before they happen.
That’s a goal I think everyone can get behind.
If you are an existing Smartvid.io customer, you can see the new tags for yourself in your org’s Sample Project (just remember to “Reset” the project to get the most up-to-date data). If you don’t have an account, reach out to us at firstname.lastname@example.org and we can get you up to speed on Vinnie’s new skills.