Have you ever thought about how bottles of soda get inspected for flaws at a bottling plant? Do you ever sit and ponder how Tesla cars can make split-second decisions without any human input? It all starts with Computer Vision.
Computer Vision is not a new technology at all but one that has been matured and can be used by many companies for a wide range of applications to automate tasks that range in complexity from identifying colours in a photo to automated crop harvesting solutions.
The technology is already powering things like Apple Face ID to recognise its owners face. It also allows barcode scanners to “see” the stripes that make up a UPC and then translate these lines into numbers. A simple way to understand Computer Vision is to think of it as the occipital lobe of the human brain which is used to process all the information we look at every day. It is not referring to the “Eyes” or cameras of the computer but rather the hardware and software used to process all this visual information into raw data.
This technology has been so commonplace in our everyday lives that we forget to admire the beauty behind how all this data is gathered and analysed so let’s look at some of the current uses.
A research group at Facebook created a technology called DeepFace which they released to their users in 2015. What DeepFace does is build a 3D Model of users photos that contain faces in them and then starts recognising pictures of those people to make tagging less of a hassle for users.
CBS has reported in the past that DeepFace is around 97% accurate which is almost on par with a human being.
Each Tesla vehicle is outfitted with multiple cameras which integrate into an advanced neural network that analyses the world around it and can make life-saving decisions on the fly.
In December of last year, the driver of a Tesla Model S fell asleep behind the wheel due to being intoxicated and instead of crashing into a barrier, his Tesla safely followed the lines on the freeway. When police officers drove in front of the car in an attempt to slow it down, the vehicle started driving slower and eventually came to a complete stop. Without Computer Vision this would not have been possible at all.
Amazon introduced its Amazon Go store early last year. These stores have no lines and no checkouts. You walk in, take the items you want and walk out. While you are in the store, their system can figure out what you took from the shelves and what you put back on the shelves and then after you walk out can send you a receipt and charge your Amazon account with no input required.
Slantrange offers drones powered by Computer Vision to provide farmers with insight into their fields. The service helps farmers identify problems with their crops such as nutrient deficiencies, dehydration, and pests. The solution is also able to estimate the farmers' total yield at harvest and can see things that aren’t easy to spot from the ground.
The drones fly over the fields and feed data to their SlantView system which is an onsite processing unit so even in locations where internet connectivity is sporadic or non-existent the data can be processed and offered to farmers in 30 minutes on a 160-acre field.
Computer Vision has come a long way from the days of being used to recognise characters on scanned documents and doing it quite badly too. With advancements in camera technology and continuous development in neural networks to process image data, it will be quite exciting to see what the future holds for this technology.