Reference · Displai
Sightcorp DeepSight Toolkit FAQ
Get answers about DeepSight Toolkit licensing, system requirements, privacy, face detection, emotion recognition, and video analysis.
When to use this
Section titled “When to use this”This article provides technical specifications, licensing information, and usage guidance for the Sightcorp DeepSight Toolkit integration within Displai. Use this when you need information about system requirements, licensing models, privacy compliance, or toolkit capabilities.
Licensing
Section titled “Licensing”Trial licenses
Section titled “Trial licenses”Trial licenses are available directly after you subscribe to the licensing server. Trial licenses have these characteristics:
- Limited in time and usage per company
- No support provided during trial period
- Allows evaluation of toolkit functionality before purchase
License types
Section titled “License types”Two license models are available:
Time-based license:
- Requires internet connection at every reboot to authenticate
- Can run offline after authentication
- Has a pre-determined availability period
Session-based license:
- Requires continuous internet connection
- Has a pre-determined availability period
Both license types require renewal after the availability period expires.
System requirements
Section titled “System requirements”Minimum specifications
Section titled “Minimum specifications”- CPU: Intel CPU with Passmark score >14k (example: i3-1220P)
- RAM: 8 GB
- Storage: 2 GB free space
Operating system support
Section titled “Operating system support”The toolkit works with x86 and x86-64 architecture:
Officially supported:
- Windows 10 (x86)
- Ubuntu 20.04 (x86-64)
Compatible (not officially supported):
- Windows (x86, x86-64)
- Linux (x86, x86-64)
32-bit versions are not available.
Privacy and compliance
Section titled “Privacy and compliance”The DeepSight Toolkit is fully compliant with privacy regulations:
- Data is collected anonymously
- No images are recorded or stored
- No uniquely identifiable data can be extracted
- Images undergo real-time conversion to data on the edge device
- Each image exists only for a few seconds during algorithmic processing
Face detection capabilities
Section titled “Face detection capabilities”Multiple people tracking
Section titled “Multiple people tracking”The software is designed for multi-person scenarios and can track multiple people simultaneously.
Face feature detection
Section titled “Face feature detection”The software is robust against:
- Beards
- Spectacles
- Sunglasses (may affect emotional classification accuracy)
Optimal conditions
Section titled “Optimal conditions”Best results occur with:
- Frontal, homogeneous natural illumination
- Clear visibility of facial features
Avoid:
- Backlighting
- Camera over-saturation
As a rule of thumb: if facial features are clearly visible in the input image, the software can interpret them.
Emotion recognition
Section titled “Emotion recognition”Detected emotions
Section titled “Detected emotions”The toolkit detects seven emotions:
- Neutral
- Anger
- Disgust
- Fear
- Happiness
- Sadness
- Surprise
Each emotion has a score between 0 and 1, with all emotions always totaling 1.
Interpreting emotion data
Section titled “Interpreting emotion data”Determining dominant emotion:
Report the highest average emotion score when that average exceeds 0.5.
Example of a person categorized as “happy”:
- happiness = 0.5 (average)
- All other emotions = 0.1 or lower (average)
When scores are ambiguous:
Categorize as “neutral” when:
- All scores are low
- Scores are too close to each other to determine a clear dominant emotion
Accuracy considerations
Section titled “Accuracy considerations”Head pose and lighting affect emotion estimation accuracy. For best results:
- Only use emotion estimations for faces within ±15 degrees of pitch and yaw (frontal faces)
Video analysis
Section titled “Video analysis”Pre-recorded video support
Section titled “Pre-recorded video support”While the toolkit is designed for real-time camera analysis, you can upload pre-existing videos for the same type of analysis. The toolkit works most effectively when combined with the DeepSight Data Studio dashboard, which displays all analyzed data.
Windows configuration
Section titled “Windows configuration”Fix stretched UI on 4K displays
Section titled “Fix stretched UI on 4K displays”If the toolkit interface appears stretched on a 4K or high-resolution screen:
- Right-click the toolkit executable and select Properties
- Select the Compatibility tab
- Check Override high DPI scaling behavior
- Under Scaling performed by, select System (Enhanced)
- Click OK
Command line arguments
Section titled “Command line arguments”Available arguments for Windows:
-s, --settings <path>— Path to the toolkit’s ini settings file-o, --show-output— Display video feed and processing result overlay in a window-r, --auto-start— Start analysis automatically when the toolkit launches
Related articles
Section titled “Related articles”- System requirements for Sightcorp integration Check the minimum and preferred processor, memory, and storage specs for hardware running the Sightcorp audience analytics integration.
- DeepSight Toolkit license types (Sightcorp) Compare DeepSight Toolkit’s time-based and session-based licenses, including when each one needs an internet connection.
- Use emotion recognition (Sightcorp) Interpret the seven emotion scores DeepSight Toolkit reports for each detected face, and get more accurate emotion results.
- Sightcorp privacy compliance Understand how Sightcorp audience analytics collects anonymous data without storing images or personally identifiable information.
Still having trouble?
Text the 24/7 line or email support.