Edge Computer Vision applications for Enterprises

Computer vision (CV) is a subdomain of the field of Artificial Intelligence that is aimed at helping machines to identify and understand content in digital video or images. Simply put, CV enabled machines to “see” the world like we humans do and use that knowledge to augment human efforts. This technology uses cameras and computers instead of the human eye to identify, track and measure targets for further image processing.

Examples of computer vision applications include self-driving cars that use multiple cameras to perceive the environment around them such as other cars, bicyclists, pedestrians, obstacles, etc., and navigate around them without any human intervention, or how your iPhone uses facial recognition to identify that it is really you to unlock your phone. Computer Vision is an active area of research and development around the world with better CV algorithms and hardware processors and hosting environments coming out almost every week!

Benefits from Computer Vision for Enterprises

As mentioned above, CV allows machines to perceive the world as humans do, but because they are machines, they can operate non-stop 24x7 and at extremely high speeds much faster than humans can, and do it objectively without any human biases or oversights, and even do it with higher accuracy. And with the advancement of technology, running CV applications on edge devices has also become inexpensive and widely adopted around the world.

CV is being used in a wide range of applications (the rest of the article is a deep dive into use cases), with more applications becoming possible very soon as the technology matures. According to a 2021 report of Verified Market Research, the AI in Computer Vision Market size was valued at USD 7 Billion in 2020 and is projected to reach USD 144 Billion by 2028, growing at a staggering CAGR of 45% from 2021 to 2028!

Enterprises can use CV to reduce costs and drive operational efficiencies by automating repetitive, time-consuming, subjective, laborious, and tedious tasks and transforming them into tasks that can be handled well by machines in an objective, non-invasive manner and run 24x7 without any need to stop at any time for any break and deployed at any scale. Enterprises can implement better Quality Control processes at a pace and scale that humans can never perform. Using CV, Enterprises can drive massive productivity gains, increase topline and accomplish more in an exponentially shorter time with significantly fewer resources.

Benefits of running CV applications on Edge devices

Edge CV is simply running CV computation on edge devices. As opposed to cloud computing, edge computing moves workloads closer to the users and closer to the source of data. Doing so gives huge advantages: real-time data processing and analysis (since data doesn’t have to travel all the way to the cloud servers and back, saving network latency concerns), minimizes network bandwidth consumption, reduces energy consumption, improves computation efficiency, and improves cybersecurity posture. Traditionally, edge devices weren’t powerful enough for CV computation and hence most of the processing must be delegated to clusters of machines in the cloud. But with the recent advances in edge technologies, the processing power has increased tremendously while being energy efficient. Additionally, Computer Vision on the edge can address any privacy concerns by guaranteeing that the data never leaves the edge device helping with changing privacy regulation landscape across the world.

CV applications by Industry

Brick and Mortar Retail

The physical retail industry is always looking to provide its customers with the best in-store experience to maximize basket size, acquire and retain customers and ultimately boost customer loyalty. CV can be used to generate critical insights and analytics in an automated fashion without the need for time-consuming techniques, employing human observers or expensive physical tracking devices, and doing it round the clock without any downtime or mistakes. Acquire repeat has needs… store owners want to know footfall, type of customers (age, gender, group size, etc.), and correlate with so they can maximize the time spent in-store and maximize basket size

• People counting: detect and track individual people in the real-time video for analysis of customer traffic. Counting people became especially popular with the COVID-19 pandemic, with stores restricting the number of people inside to prevent the spread of the pandemic. People counting can also be used to count the number of people waiting in line and notify managers to allocate more people to check out stations.

• Customer behavior analysis: track the length of time customers spend in a specific section of the store and looking at specific items (aka dwell time). CV-based applications can also be used to annotate regions of interest or boundaries and track how many people cross the boundary, how long people spend in the region of interest, and track where people are going to a visualization of how people move across the store. Such insights can also reveal peak hours and bottlenecks. Once such metrics exist for all stores, management can start comparing key metrics across different locations to identify and improve store performance.

• People heatmaps: refers to the visualization of individuals’ movement across physical spaces over time. By assigning colors to areas that see more movement relative to others and visualizing that as a contour plot, store owners can better understand points of interest, which sections see more foot traffic compared to others, and use that information for optimized product placement or mitigating bottlenecks

Manufacturing and Industrial

• Occupational safety: Manufacturing workplaces consist of moving machinery, equipment, and people and are a hotspot for accidents. CV applications can help increase worker safety, identify dangerous situations, and prevent accidents in real time while increasing productivity, and efficiency, and reducing the risk of business interruptions. For example, CV can help detect awkward human postures so they can be notified and corrected to prevent injuries.

• Detection of PPE (personal protective equipment): CV can be used to detect and monitor the usage of helmets, eye protection gear, vest detection, and more.

• Locate heavy construction equipment such as excavators, cranes, generators, or tractors, and detect when workers are operating in such dangerous areas. These automated detections can be documented, and any violations recorded so that recurring violators can be trained better or fired and safety protocols in plants or on construction sites can be updated

Public / Smart cities

Governments around the world are trying to use technology to keep people safe, detect offenders and perpetrators, and make public amenities and services easier and more convenient to use.

• License Plate Recognition (LPR): uses CV technology to read license number plates on vehicles from video feeds in real time. The latest advances in LPR technology have enabled license plate reading even for vehicles at high speeds under complex scenes and illumination. LPR is already widely in use in toll collection and parking lots management systems, detecting and identifying traffic violations, law enforcement with more applications, and higher performing, scalable and lower cost solutions being developed around the world.

• Optimized traffic lights: Instead of relying on timer-based logic, CV can be used to supercharge traffic movement by automatically detecting traffic movement, and counting vehicles in a queue so lights can be optimized dynamically to increase traffic throughput during peak as well as non-peak hours.

• Abandoned and suspicious objects detection: CV technologies can be used to identify suspicious objects that have been left behind in public spots and authorities alerted so people can be evacuated if needed and the area secured without loss of human life and property. Weapons including guns and knives can also be detected in real-time and authorities are alerted instantaneously for speedy response and loss of human life.

• Parking lot occupancy: Using CV technology for vehicle counting, parking spots can be brought online so availability in a lot is known before cars enter it, preventing the endless looping and countless minutes spent looking for a parking space, especially during rush hours. Such information could be relayed and searchable by the public and will lead to more efficient traffic flows and predictable transportation.

Physical Security and Intrusion Detection

Monitoring the physical spaces for people and personnel safety, critical infrastructure, and equipment protection have been one of the older applications of CCTV surveillance camera systems. CV has supercharged the system’s ability by autonomously detecting and responding to intrusion in real-time, without any human intervention. Modern physical security systems are intelligent enough to call for the dispatch of emergency services automatically or, in cases of confusion, request human verification.

• Person and vehicle detection: The CV algorithm running on the camera (or cloud) can automatically detect and tag all objects (people or vehicles) seen by the camera, including storing a snapshot of the people. The operator can tag one or more people or vehicles as Persons of interest (PoI) or Vehicles of Interest (VoI). The next time the person or vehicle is seen by the camera, the operator can be automatically alerted for an adequate response.

• Attribute-based search: Being able to search people by the color of clothes, gender, accessories on them, and vehicles by color, make and model makes it exponentially easy to find the people in the recorded footage instead of scrolling through days of footage.

• Search across multiple cameras: CV technology can allow people and vehicles to be searched across all cameras in a location which is invaluable in creating a forensic timeline of events for speedy investigation resolution.

• Face match: CV models can be used to match faces to determine similarity and identify if 2 pictures are of the same person. This can be used to find all instances where a given person was spotted by CCTV cameras across multiple days.

• Face blur: Smart face blur can be used to detect, and blur the faces of all non-relevant people in a security tape, protecting their privacy

Farming and Agriculture

CV plays a critical role in enabling smart farms and ranches, specifically for livestock detection, tracking, and even monitoring their health! There is increasing pressure on the limited land and resources to continue feeding the expanding world population. At the same time, the demand for sustainable livestock products, coupled with the rapidly increasing cost of labor and worker shortage has drastically increased the demand for technologies that can deliver higher operational efficiencies viz. increase productivity with a significantly smaller worker base and at a lower cost. One can use multiple static cameras or use UAVs as well. That depends on the size of the area to be monitored

Livestock monitoring solutions

• Animal counting and tracking: know where each animal is at any time, if they are in safe zones or if they need to be guided
• Animal classification: Differentiate between different types of animals on a farm and maintain a count of animals of each type
• Define regions of interest: Detect when animals cross that that zone and send alerts
• Analyzing animal activities and behavior: Identify behavior such as drinking, feeding, sleeping, birthing, and nursing and behaviors such as aggression, laziness, stress, etc.

Crop monitoring solutions

• Measurement of plant growth: by placing cameras in the right areas, growth trends can be tracked, and anomalies identified in real time
• Precision agriculture: analyze the impact of water, light, fertilizers, etc. on plant health to find optimal levels of input to achieve the highest efficiencies
• Pest/weeds/disease detection: Using CV models trained on different types of pests, weeds and occurrence of diseases, such unwanted development can be identified in real-time and humans alerted for fast response to contain its spread before a loss of output
• Sorting of crops: CV technologies can be implemented to identify the highest quality crops and produce and separate them from those that don’t meet the quality bar to better plan for shipments, locations, and markets.

Healthcare

CV has massive benefits in a wide range of healthcare applications — from patient health monitoring to environmental analysis to commercial uses such as fitness, training, and workout coaching. The benefits are that this can be done 24x7 without worrying about sleeping attendants, human oversight, and late-night labor. Best of all, the privacy of the patients can be protected through on-device processing with no visuals leaving the camera at any time.

• Fall detection: Falls are a major problem worldwide with 30% of falls causing severe injuries hence it is no surprise that fall detection is seeing rapid adoption in today’s aging population. Cameras placed in patient spaces can monitor patients’ movements including sudden changes such as falls and slippages. When a fall is detected, personnel can be notified immediately instead of relying on human input and routine check-ins to catch the mishap, leading to speedy treatment and faster recovery. Best of all, CV-based fall detection doesn’t need any setup nor do they need any expensive wearables or worry about those devices being invasive or need charging or simply forgotten.

• Posture recognition: detect the human body in different postures such as lying down, sleeping, sitting, walking, standing, etc. Posture detection can be used to identify suspicious behavior such as violence or aggression, incidents such as fall detection or injuries, and miscellaneous healthcare applications that need to measure time spent in different activities such as walking, sitting, and standing for comprehensive health and activity reporting.

• Social distancing monitoring and mask detection: with the recent pandemic most government and regulatory bodies recommended maintaining 6 feet distance and wearing masks in public spaces. CV-based technologies can easily detect human violations of these regulations and warn the people involved or automatically alert authorities.

• Fitness and online workout coaching: A surge of recent home-based workout devices rely on CV to detect posture during different workouts and recommend corrections to prevent injuries and get the most out of the workout

• Detect and blur faces in real-time: CV can be used to detect human faces and selectively blur them to protect their privacy. Such tools are invaluable for an incident response when sharing video and evidence with authorities so that the privacy of those who are not involved with the incident is protected.

Also published here.