Behavior labels per frame
Discrete behavioral state label and classifier confidence score at every frame. Standard ethogram includes locomotion, rearing, grooming, freezing, and resting.
CSV
← Capabilities
Behavior Classification
Automated classification of grooming, rearing, and user-defined behaviors
ML classifiers label behaviors frame-by-frame with temporal smoothing — replace hours of manual ethogram scoring with reproducible automated annotation.
Framelevel
Classification resolution
Weighted
Temporal smoothing
Custom
Classifier training supported
Reproducible
Deterministic scoring
The problem
Manual ethogram scoring is slow, subjective, and unreliable
Trained observers spend hours scoring behavioral videos frame by frame. Inter-rater agreement is often modest, threshold criteria drift across long scoring sessions, and expanding the ethogram to include additional behaviors multiplies scoring time linearly. The result is a bottleneck that limits sample sizes and delays analysis.
- A trained observer requires 3-5 hours to score a 30-minute behavioral session for a standard ethogram
- Inter-rater agreement (Cohen's kappa) for complex behaviors like grooming subtypes rarely exceeds 0.7
- Adding behaviors to the ethogram scales linearly — each new category requires a full re-scoring pass
The solution
ML classifiers with temporal smoothing and custom training
ConductVision applies machine learning classifiers to pose and trajectory features, assigning behavioral labels at every frame. Weighted temporal smoothing prevents frame-to-frame label flicker. Users can define and train classifiers for custom behaviors specific to their experimental model.
- Frame-by-frame behavioral labels assigned by ML classifiers trained on pose and trajectory features
- Weighted temporal smoothing eliminates single-frame label flicker without losing genuine rapid transitions
- Custom classifier training — define new behaviors, annotate examples, train, and deploy on your data
Endpoints
Ethogram-level dependent variables
Bout frequency and duration per behavior
Number of bouts, mean and total duration, and inter-bout interval for each classified behavior. Summarized per session and per user-defined time bins.
CSV
Transition matrices between behaviors
Probability of transitioning from each behavioral state to every other state. Reveals behavioral structure and sequence patterns altered by experimental manipulation.
CSVJSON
Applications
Paradigms requiring automated behavioral annotation
OCD-like repetitive behavior quantification
Score grooming bout frequency, duration, and sequential pattern in SAPAP3 knockouts or marble burying models. Temporal smoothing preserves micro-bout structure while eliminating noise.
Measures
- Grooming bout frequency
- Grooming duration
- Sequential grooming pattern
Dermatitis model grooming assessment
Quantify scratching and grooming directed at specific body regions in atopic dermatitis or pruritus models. Classify cephalocaudal grooming sequence disruption.
Measures
- Scratch bout count
- Grooming duration by region
- Cephalocaudal sequence fidelity
Stereotypy detection and classification
Identify and classify repetitive motor patterns — circling, head bobbing, route tracing — in pharmacological or genetic models. Distinguish stereotypy from normal repetitive behavior.
Measures
- Stereotypy bout frequency
- Pattern repetition index
- Stereotypy vs normal behavior ratio
Comprehensive ethogram generation
Generate a complete behavioral profile — time budget across all classified behaviors — for phenotypic comparison across genotypes, treatments, or developmental time points.
Measures
- Time budget per behavior
- Behavioral diversity index
- Activity/inactivity ratio
Compared to typical systems
How ConductVision differs
| Feature | ConductVision | Typical systems |
|---|---|---|
| Scoring speed | Real-time — seconds per session | 3-5 hours per 30-minute session |
| Reproducibility | Deterministic — identical output | Variable — kappa 0.6-0.8 |
| Custom behavior support | Train new classifiers from annotation | Fixed ethogram in most systems |
| Temporal resolution | Frame-level with smoothing | Typically binned to seconds |
| Transition analysis | Full transition matrix output | Manual compilation required |
Related capabilities
Unsupervised Discovery
Unsupervised behavioral clustering reveals structure without predefined ethograms — UMAP visualization and novel behavior flagging.
Freezing Detection
Dual-mode automated freezing scoring with TTL/serial apparatus integration for fear conditioning paradigms.
Pose Estimation
Multi-keypoint body part tracking — head, paws, tail, body center — with pretrained and custom model support.
Automate your ethogram scoring
Upload a session and see grooming, rearing, and locomotion scored automatically with frame-level resolution.
