Paradigm-specific behavioral events
Define labels, intervals and instantaneous events for the task, then verify candidate locations and boundaries.
XUANHENG · BEHAVIORAL RESEARCHDemo available
Behavior Analytics
Intelligent mouse behavior recognition
A trained model directly predicts behavior categories, segment timing and instantaneous events from video, covering 7 scenarios and 19 behavior categories. A linked timeline supports source-video review.
- 7 experimental scenarios · 19 behavior categories
- Direct model prediction of categories and timing
- Linked source-video and event-timeline review
Enlarge imageRESEARCH APPLICATIONS
Locate behavioral segments and events in video time
Analysis inputs Experimental paradigm, behavior definitions and continuous video
Temporal annotation of continuous video
Locate candidates using timelines and filters, recording overlapping intervals and manual confirmation.
Research evaluation of recognition methods
Split data by individual, distinguish negative from unannotated observations, and evaluate events and their temporal overlap.
END-TO-END VIDEO BEHAVIOR PREDICTION
7 experimental settings, 19 behaviors
Scene-specific configurations support behavioral segment and instantaneous event prediction.
Experimental settings
- EPM
- Elevated plus maze
- LH
- Learned helplessness
- NOR
- Novel object recognition
- Object
- Object-based experiments
- OFT
- Open field
- SI
- Social interaction
- Y-maze
- Y-maze
Behavioral categories
- Locomotion
- Immobility
- Freezing
- Grooming
- Sniffing
- Rearing
- Climbing
- Stretching
- Scratching
- Avoidance
- Backward movement
- Escape
- Turning
- Head turning
- Head orientation
- Head dipping
- Jumping
- Jumping up
- Jumping down
Each setting uses its own label set; not every behavior applies to every setting.
CORE CAPABILITIES
Analyze, review, and organize results
7 scenarios and 19 behavior categories
Configure labels for each experimental scenario and handle sustained segments separately from instantaneous events.
Video time linked to subject information
Retain video timestamps and associate global scene and subject-local information to locate replayable intervals.
Direct prediction of behavior and timing
A temporal model combines scene, subject-local information and localization reliability to predict categories and temporal boundaries directly.
Interval and instantaneous events
Support overlapping behavioral segments and instantaneous events with their respective timing boundaries.
Timeline and video review
Filter candidates by category, event type and time, replay the video, and save confirmed results.
Annotation and research evaluation
Distinguish explicit negatives from unobserved/unannotated data, split by individual, and evaluate using measures such as temporal overlap.
Software in action
Actual interfaces for parameter configuration, result filtering, and review; original screenshots retain their Chinese labels.
Candidate filtering and time-range review
Feature example
Enlarge imageRESEARCH OUTPUTS
Results and research interpretation
Measured examples from Behavior Analytics, preserving original figures and quality information.
Summary of filtered event candidates
Result example
Enlarge imageREPORT EXAMPLES
Research reports for your records
Rendered from PDFs actually exported by the software, retaining original figures, labels, and quality notes.
Behavior-candidate summary and timeline
Report example
Enlarge imageEvent timing and detailed records
Report example
Enlarge imageANALYSIS WORKFLOW
From data preparation to research outputs
- 01
Define the task and behaviors
Select the paradigm and check definitions for behavior categories, sustained intervals and instantaneous events.
- 02
Import video and configure analysis
Link continuous videos to task parameters and inspect source-video and subject information.
- 03
Inspect candidates and timing
Locate behavioral segments and instantaneous-event candidates using the event timeline and filters.
- 04
Confirm against the source video
Check categories, boundaries and overlaps, then save annotation and confirmation records.
- 05
Evaluate and export confirmed results
Review annotation and evaluation information as required and export confirmed analysis data.
DELIVERABLES
Your analysis outputs
- Behavioral segments and event candidates
- Timeline and video review
- Annotation and evaluation data
- Export of confirmed results
BEFORE ANALYSIS
What to confirm before you start
- The 7 scenarios use their corresponding label sets; not every scenario includes all 19 behavior categories.
- Recognition performance must be evaluated on the target experimental data. Candidate scores are not accuracy estimates.
- Check model-output categories and temporal boundaries against video and obtain manual confirmation before research use.
FREQUENTLY ASKED QUESTIONS
Before you begin
Does end-to-end prediction remove the need for review?
No. End-to-end means the model directly predicts categories and timing, rather than creating labels from trajectory rules. Research use still requires source-video review; coverage is not an accuracy guarantee.
Does an analysis score represent recognition accuracy?
No. Scores support ranking and filtering. Evaluate recognition against manual annotations that follow the experimental definitions, using appropriate evaluation measures.
How should I choose between behavior recognition and tracking?
Choose Behavior Studio for distance, speed and zone occupancy; explore Dense Tracking Studio for body-surface motion and interactions; choose Behavior Analytics for behavior categories, occurrence times and boundaries.
Find the right approach for your research
Tell us your experiment, recording angle, and target measurements to arrange a Behavior Analytics demo and sample assessment.
For research use only. Figure captions describe their sources and scope. Examples do not guarantee accuracy, diagnostic suitability, or research conclusions.