Yolo = You Only Look Once. Yolo v4 is an object-detector that can identify objects in pictures and video.
In this example, you can see how Yolo detects 7 people, 1 bike, 2 cars, 1 truck and 1 traffic light.
I am amazed over where we are technological. All of this was detected in an instance. Yolo did not run for a couple of hours or minutes even. This was done like a snap of your fingers.
Personally I count: 19 people, 7 bikes, 3 car, 3 trucks, 11 traffic lights ( took me more than one look)
I don’t quite understand the difference between cars and trucks in this image. I also don’t understand how the blue car on the left is detected and only one of the cars waiting for green. Both have the same amount of car shown. The blue is however closer giving it more pixels. + Of all the traffic light Yolo could detect it sees one without any light.
You can also extract the identified objects from the image.
Here I have created clusters of the identified objects. It provides a segmented insight into the original image.
Found this little hiccup while making the clusters.
Yolo v.4 as a design tool
One example: I think Yolo could provide a new way of gathering information from design probes. In a lot of traditional probe kits a disposable camera have been included.
Today most people carry a phone with a camera and an internet connection. This enables designers to receive images on the go from the ‘probers’. It could be interesting to use Yolo to interpret the incoming photos.
Probe photo -> Yolo -> Yolo output -> designers
Will information be lost? sure. But it might be easier to compare different shots or provide new insights.