Gaze Detection

Class: GazeBlockV1

Source: inference.core.workflows.core_steps.models.foundation.gaze.v1.GazeBlockV1

Run L2CS Gaze detection model on faces in images.

This block can: 1. Detect faces in images and estimate their gaze direction 2. Estimate gaze direction on pre-cropped face images

The gaze direction is represented by yaw and pitch angles in degrees.

Type identifier

Use the following identifier in step "type" field: roboflow_core/gaze@v1to add the block as as step in your workflow.

Properties

Name	Type	Description	Refs
name	str	Enter a unique identifier for this step..	❌
do_run_face_detection	bool	Whether to run face detection. Set to False if input images are pre-cropped face images..	✅

The Refs column marks possibility to parametrise the property with dynamic values available in workflow runtime. See Bindings for more info.

Available Connections

Compatible Blocks

Check what blocks you can connect to Gaze Detection in version v1.

• inputs: VLM as Detector, Corner Visualization, Camera Focus, Polygon Zone Visualization, Circle Visualization, Roboflow Dataset Upload, PTZ Tracking (ONVIF).md), Triangle Visualization, Stability AI Inpainting, Roboflow Custom Metadata, Classification Label Visualization, Bounding Box Visualization, Depth Estimation, SIFT, Image Convert Grayscale, Halo Visualization, Grid Visualization, Dynamic Zone, Email Notification, Polygon Visualization, Absolute Static Crop, Slack Notification, Dot Visualization, Color Visualization, Label Visualization, Stability AI Outpainting, Detections Consensus, Crop Visualization, Identify Outliers, Perspective Correction, Stability AI Image Generation, Image Slicer, Model Monitoring Inference Aggregator, Image Threshold, Image Preprocessing, Model Comparison Visualization, VLM as Classifier, SIFT Comparison, Dynamic Crop, Stitch Images, Image Contours, Mask Visualization, Identify Changes, JSON Parser, Webhook Sink, Twilio SMS Notification, Pixelate Visualization, Roboflow Dataset Upload, SIFT Comparison, Camera Calibration, Reference Path Visualization, Image Blur, Line Counter Visualization, Local File Sink, VLM as Classifier, Background Color Visualization, Image Slicer, Keypoint Visualization, Blur Visualization, Ellipse Visualization, Trace Visualization, VLM as Detector, Relative Static Crop
• outputs: Corner Visualization, Google Gemini, Detections Classes Replacement, OpenAI, Roboflow Dataset Upload, Circle Visualization, PTZ Tracking (ONVIF).md), Triangle Visualization, Roboflow Custom Metadata, Detections Transformation, Bounding Box Visualization, Florence-2 Model, Detections Merge, Distance Measurement, Template Matching, Dynamic Zone, Dot Visualization, Label Visualization, Color Visualization, Detections Consensus, Crop Visualization, Detections Filter, Model Monitoring Inference Aggregator, Detection Offset, OpenAI, Model Comparison Visualization, Dynamic Crop, Webhook Sink, Florence-2 Model, Segment Anything 2 Model, Pixelate Visualization, Llama 3.2 Vision, Roboflow Dataset Upload, Byte Tracker, Camera Calibration, Line Counter Visualization, Anthropic Claude, Background Color Visualization, Velocity, Keypoint Visualization, Blur Visualization, Ellipse Visualization, Trace Visualization

Input and Output Bindings

The available connections depend on its binding kinds. Check what binding kinds Gaze Detection in version v1 has.

Bindings

• input

  • images (image): The image to infer on..
  • do_run_face_detection (boolean): Whether to run face detection. Set to False if input images are pre-cropped face images..

• output

  • face_predictions (keypoint_detection_prediction): Prediction with detected bounding boxes and detected keypoints in form of sv.Detections(...) object.
  • yaw_degrees (float): Float value.
  • pitch_degrees (float): Float value.

Example JSON definition of step Gaze Detection in version v1

{
    "name": "<your_step_name_here>",
    "type": "roboflow_core/gaze@v1",
    "images": "$inputs.image",
    "do_run_face_detection": "<block_does_not_provide_example>"
}