Processing optical person attributes

The AI Person Attributes App and the underlying neural network AI Person Attributes have been especially developed for determining  the hue of clothing on the upper and lower body of persons.

In combination with Dallmeier client software, this creates an efficient solution that supports finding of corresponding persons in extensive video recordings.

This significantly speeds up targeted evaluation and saves valuable time.

AI Person Attributes App
The AI Person Attributes App allows users to adjust the analysis interval of the neural network to suit the scene in question. For example, a shorter interval can be selected for smaller scenes with fast-moving persons in order to achieve optimal results. After each analysis, the app compiles the recorded attributes and assigns them to the recognized persons. The resulting data packets are then transmitted together with the video stream to the Dallmeier recording system, where they are immediately available for evaluation with Dallmeier client software.

AI Person Attributes Network
The neural AI Person Attributes network uses the object classification of the camera and determines the dominant hue of a person's clothing, taking color nuances into account. For this purpose, it first segments the upper and lower body for efficient color determination, while leaving out uncovered arms, legs, and faces. This targeted reduction also helps to prevent potential discriminatory misuse of the system. Regardless of this, under favorable conditions, the network recognizes whether the person is wearing headgear, a backpack, or a bag. Finally, all captured attributes are collected at the set interval and are immediately available for further processing in the AI Person Attributes App.

Client Software
For small to medium-sized systems, SeMSy® Compact is suitable for evaluating the analysis results and associated recordings from a camera. After defining relevant attributes using the SmartFinder function, potentially matching persons are displayed with preview images. These allow direct access to the associated recordings with just one click, which significantly speeds up the process of finding relevant events. Instead of completely reviewing long sequences or all recognized persons, it is sufficient to evaluate a few specific events.

Client System
Larger Hemisphere® SeMSy® systems will support the evaluation of person attributes using the SmartFinder function in one of the next versions. This will offer the same time-saving evaluation functionality that is already available with SeMSy® Compact.

Compatibility
The AI Person Attributes app and the associated neural network can be used with Domera® RDF6400DN, RDF6800DN, and SDF6800DN cameras, as well as with all Panomera® S/W/V 4 and 8 cameras that support AI Analytics.

Licensing
After entering the free license code on the camera, the AI Person Attributes App is activated. The neural AI Person Attributes network must be installed additionally after licensing.

Notes on using the App
The AI Person Attributes app and the associated neural network can be used simultaneously with the EdgeAnalytics AI Essential apps, such as Line Crossing or Intrusion Detection.

Notes on using the network
The neural AI Person Attributes network requires visual conditions comparable to human perception for person recognition and approximate color determination. Under unfavorable lighting conditions, low contrast, or when persons are heavily obscured, analysis results may therefore vary.

Compatible Cameras

   
Panomera® V8 AreaPanomera® V8 Area
Multifocal sensor system, 30 fps, H.265, day/night, UWDR, field of view horizontal 180° and vertical 90°, resolution 125 px/m up to 40 m, Panomera® View Package (included), AI EdgeAnalytics (optional), variable Mountera® mounting system
Panomera® S8 Area SeriesPanomera® S8 Area Series
Multifocal sensor system, 30 fps, H.265, day/night, UWDR, field of view horizontal 22°-90°, resolution 125 px/m up to 212-71 m, Deep AI object classification, EdgeAnalytics AI Apps, variable Mountera® mounting system
Panomera® S4 Area SeriesPanomera® S4 Area Series
Multifocal sensor system, 30 fps, H.265, day/night, UWDR, field of view horizontal 15°-30°, resolution 125 px/m up to 194-97 m, Deep AI object classification, EdgeAnalytics AI Apps, variable Mountera® mounting system
Panomera® S8 Stadium SeriesPanomera® S8 Stadium Series
Multifocal sensor system, 66 MP, 30 fps, H.265, day/night, UWDR, field of view horizontal 13°-61°, resolution 125 px/m up to 325-102 m, Deep AI object classification, EdgeAnalytics AI Apps, variable Mountera® mounting system
Panomera® S4 Stadium SeriesPanomera® S4 Stadium Series
Multifocal sensor system, 33 MP, 30 fps, H.265, day/night, UWDR, field of view horizontal 12°-41°, resolution 125 px/m up to 325-103 m, Deep AI object classification, EdgeAnalytics AI Apps, variable Mountera® mounting system
Panomera® W8 AreaPanomera® W8 Area
Multifocal sensor system, 30 fps, H.265, day/night, UWDR, field of view horizontal 360° and vertical 90°, resolution 125 px/m up to 17 m, Deep AI object classification, EdgeAnalytics AI Apps, variable Mountera® mounting system
Panomera® W4 AreaPanomera® W4 Area
Multifocal sensor system, 30 fps, H.265, day/night, UWDR, field of view horizontal 180° and vertical 90°, resolution 125 px/m up to 17 m, Deep AI object classification, EdgeAnalytics AI Apps, variable Mountera® mounting system
Camera Series RDF6000DNCamera Series RDF6000DN
High definition camera, 4/8 MP, H.265, day/night, WDR/HDR, PoE, Deep AI object classification, EdgeAnalytics AI Apps, motor-driven varifocal lens, motorized 3-axis adjustment (RPoD), variable Domera® mounting system
SDF6800DNSDF6800DN
Ultra high definition camera, 8 MP, H.265, day/night, WDR, PoE, Deep AI object classification, EdgeAnalytics AI Apps, fisheye lens, 360° panoramic view, motorized fine alignment, variable Domera® mounting system