Load videos or images of worker sites to detect the usage of PPE. Label images using automated annotation to specify what PPE looks like in your environment, and easily train a deep neural network in one click. Take a look below at what teams on V7 built for PPE detection.
Hover over the image below to see what the AI identifies
According to NIOSH, over 2,000 work-related injuries occur every day in the US which could be prevented through the use of PPE. Deep learning has enabled the detection of protective equipment across multiple domains, as long as training data is varied and representative enough. To leverage V7 for PPE detection, start by uploading your domain's video footage or images displaying protective equipment such as hardhats, safety glasses, or face masks. Label them using V7's annotation tools, and then train a model to adapt to a new camera deployment or set of equipment.
PPE comes in many forms, items that can be reliably spotted with machine learning are:
The best ML models start to learn objects across domains after they've seen 1,000 examples, but you can start training with as little as 30 on V7. What's important is that you fine-tune your model upon new deployments. This means uploading a small sample of data from the new factory, construction site, or laboratory where the PPE detection is happening, so that the AI model adapts to what the camera quality, background, and subjects look like. There are no limitations when it comes to lighting or angles to worry about, however make sure your object's sizes are at least 1% of the image's width. If they aren't, consider labelling a taxonomy level above that (for example rather than "safety glasses" try "face with safety glasses".
AI performance shouldn't stop at the first training session.
Sync your product with a dataset and continually improve your AI performance by turning its output into new ground truth. Continual learning will allow your models to surpass 95% accuracy barriers and reach 99% and above as more data from your use case or product is supervised by humans, and learnt by your AI.