Human bounding-box detection — frame-level center X/Y and kinematic parameters.
2D-AI Human-Dog Interaction Behavior Analysis System
BehaviorAtlas Social Analyzer Dog — single overhead 2D multi-target tracking: human, dog, and object classes with tiered representation (dog skeleton keypoints; human/object bounding boxes). Frame-level coordinates, travel distance, and average speed.
// tracking stream
[frame 1240] HUMAN bbox center (312, 186) px
3
Target classes
2D
Plane tracking
Frame
Metrics
Research Background
Human-dog interaction is hard to quantify objectively — single-camera 2D tracking makes cross-species social behavior measurable and comparable.
Human-dog interaction research, pet assessment, and working-dog training have relied on manual observation — lacking frame-level, cross-sample comparable spatial data. Real scenes include people, dogs, and toys/objects requiring stable IDs and continuous trajectories through occlusion and crossing motion.
BehaviorAtlas Social Analyzer Dog is built for human-dog interaction: single overhead camera capture with AI identifying three target classes — humans and objects via bounding-box centers, dogs via 2D body keypoints. Automatic speed, distance, and four-channel parameter charts — model test preview, sample grouping, and export-ready outputs.
Three Target Classes
Synchronized detection of human, dog, and object (toy) targets in typical interaction arenas.
Tiered Representation
Dog: 2D skeleton keypoints; human/object: bounding box + center trajectory.
Quantified Metrics
Frame-level X/Y, travel distance, average speed — with real-world unit conversion.
Detection Model: Three Classes · Two Tiers
In one overhead view, human, dog, and object are detected together — dogs via 2D skeleton keypoints; humans and objects via bounding-box center tracking.
Dog 2D skeleton keypoint tracking — selectable body points in model test, output to 2Dskeleton h5.
Toy/object bounding boxes — center trajectories for interaction distance and spatial analysis.
Core Software
Integrated 2D human-dog analysis — project setup, model testing, batch analysis, database preview and export.

BehaviorAtlas Social Analyzer Dog
Project setup · 2D tracking · grouped analysis · parameter export
Built for 2D human-dog interaction analysis. New/open projects (.basda config), sample import, auto/custom grouping; model test preview before batch analysis — outputs 2D skeleton, bounding boxes, and kinematics to h5. Database module: frame preview, four parameter charts, and exportable parameter tables.
Software Functions
Social Analyzer Dog — project management, model testing, sample grouping, 2D tracking, and database export.

Project & Sample Management
New/open .basda projects, import .avi/.mp4 samples — data/ raw video and BeAOutputs/ structured outputs.

Smart Sample Grouping
Auto-grouping by naming convention (rec-2-group-xxx) or custom treatment/control groups.

Model Test Preview
Sample N frames before full runs to verify model match — select body points and apply model.
2D Multi-target Tracking
Dog 2D skeleton keypoints plus human/object bounding boxes — overlay preview in results.

Single-sample Data Preview
Raw video, frame coordinate tables, and four parameter charts (body x/y, target center x/y).

Parameter Table & Unit Conversion
Target average speed and total distance tables — real-world unit conversion with CSV/TXT export.
2D Analysis Workflow
New project → model test → batch analysis → result review → database export
- Project structure · .basda config
New Project & Sample Import
Set project name and path, import .avi/.mp4 videos — creates data/ folder and BA_SocialAnalyzerDog.basda config.
- Preview · select · apply
Model Test
Select model and body points, sample N frames (default 1000) for preview — apply model when matched.
- Batch 2D tracking
Run Analysis
Run full analysis on selected samples — outputs 2Dskeleton, Boundingboxes, AdvancedParameters to BeAOutputs/ h5.
- Box + skeleton overlay
Result Review
Scrub timeline to review body points and bounding box tracking frame by frame.
- Charts · tables
Database & Export
Frame coordinates, four parameter charts, speed/distance tables — export video, charts, with scale conversion.
Analysis Outputs
From 2D skeleton and bounding-box trajectories to parameter charts and tables — export-ready interaction metrics.
2D Skeleton & Bounding Boxes
Dog body keypoint coordinates per frame; human/object bounding-box center trajectories — overlay video preview.
Four-channel Parameter Charts
Body-point x/y and target-center x/y — four frame-level parameter curves in pixels, exportable as images.
Kinematics Parameter Table
Per-target average speed (pixels/frame) and total distance (pixels) — convert to physical units with room scale.
Grouped Statistics & Export
Group-level parameter summaries — CSV/TXT table export, batch archive of videos and parameter charts.
Applications
Human-dog interaction research, pet assessment, and working-dog training — single-camera 2D quantification with easy deployment.
Cross-species Social Research
Objective quantification for ASD canine models, interaction mechanisms, and affective dynamics research.
Pet Behavior Assessment
Activity range, approach trends, and movement intensity in home interaction — welfare-related metrics.
Working Dog Training
Single overhead deployment in training grounds — human-dog distance, following, and synchronized movement.
Toy-mediated Interaction
Toy as third target — joint attention, competition, and cooperation in three-agent social events.
Learn More
Contact us to learn more about objective quantification of human-dog interaction and cross-species social analysis.