User Guide

• Data Preparation
  • Directory Structure
  • Directory Tree and Experimental Conditions
  • Key Point Configuration
• Social behavior classification using a pre-trained model
  • Step 0: Prepare the data and model
  • Step 1: Annotate behaviors
  • Example: Adapting Keypoints from a New Dataset
  • Advanced: Fine-tuning a classifier
  • References
• Social behavior discovery using a pre-trained model
  • Step 0: Prepare the data and model
  • Step 1: Embedding
  • Step 2: HMM fitting
  • Step 3: Prototype selection
  • References
• Dimensionality Reduction
• Self-Supervised Tasks
  • Introduction: Why Self-Supervision?
  • The Four Core Self-Supervised Tasks
  • Group Consistency
  • Temporal Order
  • Temporal Shift
  • Temporal Warp
  • Summary Table
  • Practical Notes
  • References and Further Reading

Advanced

• Model training using self-supervised learning
  • Step 1: Prepare the data
  • Step 2: Train the model
  • [OPTIONAL] Step 3: Export embedding model
  • References
• Data augmentation
  • Available augmentation techniques
  • Usage examples
• Fine tuning a classification model on a custom dataset
  • Step 1: Load the dataset
  • Step 2: Fine-tune the model
  • References
• Calibrating the prototype selection process
• Annotating new data using selected prototypes
  • Recommended Approach: Train a LISBET Classifier on Prototypes
  • Alternative: Using Cached HMMs
  • References