Scoring and Interactive Graphs

Summary of Changes

• Add a new page for manually scoring animal behavior
• Add a new interactive graphing tool for frame-by-frame data inspection
• Add a new graph type for visualizing the differences between groups
• Several improvements to the summary CSV

Manual Behavior Scoring

This update adds a new page to the Analysis tab called Scoring. This page provides a user interface for frame-by-frame scoring of arbitrary behaviors. You can create a list of behaviors you’d like to score, then use keyboard shortcuts to navigate through the video, labeling behaviors. You can load the list of behaviors from another recording to ensure all scored recordings are consistent. The score-table is saved alongside the recording and will be imported/exported using the batch transfer system.

This update does not provide automatic scoring. That will be coming early 2025 and this update lays the groundwork for such a system.

Interactive Time Series

Another big feature provided by this update is the new interactive graphing tool. This is located in the Visualize tab in a page called Time Series. This tool allows you to watch how any measurement made by our software mutates over time. You can select multiple recordings and compare them, even within the same chart. With this, you’ll be able to unlock the full potential of your data.

Cluster Heatmap Graph

Another visualization tool added by this update is located in the Visualize > Cluster Heatmap page. This visualization provides researchers with powerful insights by revealing latent structures, highlighting similarities and differences across groups, and pinpointing significant features. It aids in understanding how well group identities align with data patterns, uncovering relationships that can inform future analyses and experimental planning.

Group-Level Cluster

This mode visualizes mean Z-scores across all feature columns, complete with signifiance markers (e.g. *, **, ***). It incorporates ANOVA-based statistical testing with Bonferroni-corrected p-values, highlighting significant features across different groups. Hierarchical clustering orders features, making group comparisons intuitive and clear.

Individual-Level Cluster

This mode presents Z-scores for all feature columns at the recording level. Recordings are sorted using hierarchical clustering, independent of group identity, to uncover latent cluster structures within the dataset. Once clustering is applied, the group identity is recovered and displayed, allowing users to assess whether these underlying structures align with predefined groupings. This approach is valuable for identifying hidden patterns and verifying if natural data clusters correspond with existing group labels.

Summary CSV Improvements

This update also adds some highly-requested changes to how summaries are generated. With this update, creating a new directory for each analysis is no longer required, though doing so is still often desireable.

Filename Change

The summary.csv file will now have a date and time in the filename, allowing you to easily reuse a directory for analysis outputs. All newly generated CSVs will have a filename in the format: summary_YYYY-MM-DD_HH-MM-SS.csv.

Name Column

The first column – which previously had no name – is now called name and contains the animal’s name.

Column Order

The order of columns has been altered to make the summary more easy to interact with in e.g. Excel. Immediately following the name column (the first column) are all of the variables entered in when the capture was taken. These columns were previously the last columns in the summary.