Computer vision

Computer vision is a technology that uses artificial intelligence and machine learning to detect specific objects or activities and derive information from video footage and images. A computer vision algorithm can be trained to detect road signs, analyze weather conditions, or classify images, for example.

A computer vision model requires large datasets of labeled videos or images to identify the patterns and rules that it will later use to fulfill its purpose.

For instance, to recognize a crack in a wall, a computer vision algorithm has to be fed a large dataset of images of walls with and without cracks. The algorithm will then “learn” the difference between a wall with and without cracks.

Here are the ten practical applications of computer vision across industries:

• Identifying weather conditions in the dashcam footage
• Using facial recognition for authorization and security purposes
• Enabling autonomous driving by analyzing the road: recognizing road signs, detecting lane markers and potholes, identifying pedestrians
• Analyzing security camera footage in real time (e.g., detecting when doors open or close)
• Detecting defects in items on the production line in real time
• Monitoring livestock behavior and identifying animals with potential health problems
• Automating UI interface testing
• Assessing and updating parking lot occupancy in real time
• Analyzing traffic flow using CCTV footage
• Detecting queues in retail stores and notifying staff that new checkouts should be open once a threshold is reached

Here are four examples of how computer vision is transforming retail:

• No-checkout retail. Instead of manually scanning barcodes at the checkout, customers can pick up items while a computer vision solution identifies every object using camera footage
• Inventory management. Computer vision solutions can detect inventory gaps in real time by identifying empty shelves or misplaced items. They can also be trained to detect damaged packaging or incorrect pricing
• Store layout. Computer vision can power heat mapping of the store layout, assigning a specific color to each area based on foot traffic. Such heat maps allow the enterprise to better understand customer behavior and improve store layout accordingly
• Virtual mirrors. Virtual mirrors demonstrate how an item of apparel or accessory would look on the customer without them having to try it on. They are made possible by a combination of computer vision and augmented reality technologies