Object Detection in Video: Capture More than Images

Object Detection in Video: Capture More than Images

The amount of video is growing rapidly across industries. A recent IDC report projects that 79.4 zettabytes of data will be created by connected Internet of things (IoT) devices by 2025, mostly generated by video applications.1

Harnessing the full potential of this content deluge is no small task. Manual video analysis is a slow, labor-intensive process that is prone to errors and often not feasible due to lack of resources.

Failure to fully optimize content or effectively monitor and respond to live video can result in missed opportunities for a variety of industries. Below is a sampling of how object detection in video and overall video analytics technology is impacting specific industries:

  • Retail – To avoid losing shoppers to the competition in both physical and digital retail spaces, creating a superior in-store experience is critical. A lack of insight into customer behaviors, crowded checkout lines and on-the-shelf inventory may inhibit a retailer’s ability to drive sales by refining in-store operations, improving staffing and growing foot traffic. Retailers must also take steps to help mitigate the estimated USD 50.6 billion in losses2 attributed to shoplifting, employee theft and other errors across the industry.
  • Government – When it comes to government challenges like situational awareness and forensic investigations, the explosion in the amount of video is driving public and federal agencies to develop Artificial Intelligence (AI) video solutions that can provide secure environments to protect our national interest against our adversaries, safe environments to serve public needs for recreation, safe transportation, import/export of goods and services as well as worker safety.
  • Travel and transportation – More passengers means more revenue, and not only from ticket sales. Counting people, their speed and direction can provide usable insight into foot traffic patterns, gate operations and customer movement through public areas and retail stores. These insights can help transportation authorities develop better ways to manage crowd safety and the operational efficiency and profitability of retail areas.
  • Energy, Utilities, Mining, Oil and Gas – Supporting the safety of workers and the continued operation of key infrastructure constant monitoring of critical areas. Damage to critical equipment can lead not only to service downtime, but also dangerous situations.
  • Agriculture – Growers and producers must remain diligent about detecting hazardous conditions for workers – both in the field and in storage structures or warehouses – and ensuring the proper use of protective gear, which can be challenging for farms that cover a wide geography. Also, the inability to track livestock movement and behavior may have negative impacts on food quality and production.
  • Media – As competition grows across platforms, keeping a viewer’s attention may require more than simply producing the best content. Media companies must be able to categorize and organize content so that it can be matched (“search and discover”) to a viewer’s specific needs or provide context to an event, delivering the right video at the right time. At the same time, media companies need solutions to help them identify and filter out inappropriate content to comply with sponsor contract terms.

Gain insights through automatic object detection in video

Many organizations are now exploring the use of AI to rapidly analyze live and archived video to gain a deeper understanding of customers, infrastructure and operations.

By using AI models that are customized for the unique needs of a business or industry, these solutions help drive benefits such as:

  • Enhancing operational insights – Measuring human and vehicle traffic in a specific area to identify movement patterns, dwell times, and common areas of congestion.
  • Enabling proactive responses – Delivering near real-time alerts when potentially problematic circumstances or events are captured on video, such as a checkout queue becoming too long in a store, someone entering a restricted area, a railway platform becoming overly crowded or a suspicious item left in a public zone.
  • Supporting worker wellbeing – Identifying potentially dangerous situations, including safety or protective equipment not being worn properly, and alerting personnel.
  • Identifying and removing inappropriate content – Monitoring people, behaviors and logos in real-world and digital spaces so that unwanted and inappropriate content can be flagged and addressed.
  • Alerting law enforcement or security personnel – Monitoring perimeters for unauthorized entry, appearance of weapons, scaling of fences by humans versus animals, identifying formation of crowds/protest groups in public areas.
  • Complying with mandates – Pandemic situations have heightened requirements for specific personal protective equipment (PPE) that must be required and enforced for employees, customers, and the public to varying degrees.

IBM Video Analytics is designed to work with virtually any existing video cameras (often existing security cameras that are already in place) and video management system, applying AI models that are tailored for the unique needs of your business or industry. The solution comes with some models out of the box for object detection in video, but is also integrated with IBM Maximo Visual Inspection to help you create new AI models, enabling more customized search and analysis without the need to add data scientists to your staff.

The result is an AI-powered, data driven, computer vision solution that helps unlock deeper insights into the most critical factors of your business and supports your ability to make faster, better decisions. Operations teams can scale the use of this solution with their needs, starting with a few cameras and then scaling up as needs and use cases for automated video analytics grow.

Cameras capture video, but IBM Video Analytics helps you capture value. To learn more or request a demonstration, contact us today.

1 Worldwide Global DataSphere IoT Device and Data Forecast, 2019–2023, Carrie MacGillivray and David Reinsel, IDC
2 National Retail Security Survey, 2019, National Retail Federation

Subscribe to updates