Introduction
Activity Recognition is another addition to VIDIZMO's array of AI insights. Its implementation lets you identify three activities in your videos: robbery, shopping, and trespassing. The ability to detect these activities offers an easier way to analyze your videos for specific events. The AI model utilized by VIDIZMO can also recognize these activities regardless of the video quality, environment, or the people in the videos.
Concept
VIDIZMO uses a video classification model to perform Activity Recognition. When a video is processed for activity insights, the model scans the entire video by analyzing 16 frames consecutively till it reaches the end. If the model identifies any activity occurring within those frames, it assigns a class to those frames. These classes (or activity types) are robbery, trespassing, and shopping.
Videos can be processed for activity insights after configuring Activity Recognition from your Portal's App Settings. Activity Recognition is available as a Detection Type on the VIDIZMO Indexer App. Once enabled, you can decide which activity types will be available for detection on your Portal. You must add at least one activity type to the Activity Recognition field.
Only 'Video' will be available as a Media or Evidence type for processing if only Activity Recognition is added to the Detection Types. This is because Activity Recognition only works on videos. Other types, such as 'Images' or 'Documents,' will only become selectable if other Detection Types, such as Face or Optical Character Recognition, are added.
Insights for all activity types are shown on the processed video's playback page. Detected activities can also be viewed when that video is opened in Studio Space. The detected activities' start and end timestamps and duration are displayed in Studio Space, allowing the user to navigate to the relevant parts of the video.
Use Cases
Analyzing Security Footage
Activity recognition aids the user in identifying specific activities throughout their video. It can substantially decrease the time it takes for a user to analyze the footage of an incident that involves these activities, such as robbery or trespassing. Law enforcement agencies can benefit from activity recognition as it helps them perform analysis of video evidence much more quickly. Without having to sit through and look for an event, officers can quickly jump to the events detected by the model by clicking the activity insights or detections.
Improving Marketing Strategies
Marketers or retailers can utilize activity recognition to analyze customer behavior and develop profitable strategies. They can use Shopping activity insights to identify which products were bought the most or which aisle had the most customers in their store. This analysis can help them improve their marketing strategies, such as visual merchandising.
Up Next
To learn how to enable activity recognition in the VIDIZMO portal, refer to the article "How to Generate Activity Insights in VIDIZMO."