Skip to content

Batch Output Map

The Batch Processing Wizard keeps batch-level summaries at the top of the output folder and places each video's detailed results inside videos/.

If you are unsure where to begin, open batch_results.xlsx.

Start with the result you need

Research question Start here
Did every selected analysis run? analysis_coverage_table.csv or the Analysis_Coverage workbook sheet
What happened in each video? Group_Video_Summary in batch_results.xlsx
How far did each animal travel? metrics_summary_by_track.csv for the video
What are the individual behavior bouts? <video>_detailed_bouts.csv
How many bouts and how much behavior time were recorded? <video>_summary.csv
How long did each animal remain in an arena ROI? <video>_roi_exclusive_per_track.csv
What were the individual ROI visits? <video>_roi_exclusive_dwell_events.csv
Did an animal interact with an object? <video>_object_interactions_per_track.csv
What were the individual object visits? <video>_object_interactions_dwell_events.csv
Where can I resume an unfinished manual review? The video's bout_review_workspace/
Where are behavior-review tables and figures? The newest bout_review_exports/IntegraPose_bout_review_<timestamp>/Behavior_Bouts/
Where are ROI-review tables and figures? The newest review export's ROI_Bouts/
Where are object-review tables and figures? The newest review export's Object_Interactions/
How closely did predictions match the completed review? prediction_vs_review_scores.csv in the applicable review category
Which behavior classes required the most correction? Behavior_Bouts/Tables/behavior_correction_metrics.csv
What group comparisons were performed? group_stats/ or the statistics sheets in batch_results.xlsx
Which figures were produced? figures/figure_manifest.csv
Which optional-analysis tables were produced? module_tables/module_table_index.csv
Which settings and files belong to one video? That video's run_manifest.json

Folder layout

The exact set of files depends on the selected model, ROIs, object settings, optional analyses, figure settings, and review policy.

chosen-output-folder/
|-- batch_results.xlsx
|-- batch_session.json
|-- batch_results_status.json
|-- analysis_coverage_table.csv
|-- group_stats/
|   |-- group_stats_overview.csv
|   |-- group_pairwise_tests.csv
|   |-- group_effect_sizes.csv
|   `-- group_kpss_mixed_effects.csv
|-- figures/
|   |-- figure_manifest.csv
|   |-- assay_figure_index_<preset>.csv
|   `-- selected PNG and SVG figures
|-- module_tables/
|   |-- module_file_index.csv
|   |-- module_table_index.csv
|   `-- consolidated optional-analysis CSV files
`-- videos/
    `-- <video-id>_<video-name>/
        |-- inference/
        |   `-- infer/
        |       |-- labels/
        |       |-- inference_metadata.json
        |       |-- metrics.csv
        |       |-- metrics_summary_by_track.csv
        |       `-- metrics_summary_by_frame.csv
        `-- analytics/
            |-- <video>_detailed_bouts.csv
            |-- <video>_summary.csv
            |-- <video>_analytics_dashboard.png
            |-- run_manifest.json
            |-- bout_review_workspace/
            |   |-- IntegraPose_bout_review.sqlite3
            |   `-- last_review_status.json
            |-- bout_review_exports/
            |   `-- IntegraPose_bout_review_<timestamp>/
            |       |-- Behavior_Bouts/
            |       |-- ROI_Bouts/
            |       |-- Object_Interactions/
            |       `-- Shared_Audit/
            `-- selected ROI, object, and optional-analysis files

If a name such as infer already exists, IntegraPose uses the next available run name instead of replacing it.

Batch-level files

batch_results.xlsx

This is the main batch workbook. It brings the most useful tables together in one place.

Sheet Contents
Keypoint_Estimation Consolidated pose or detection rows when available
Kinematic_Outputs Consolidated movement measurements
Bout_ROI_Metrics Consolidated behavior bouts and ROI context
Group_Video_Summary One summary row per video
Stats_Omnibus Overall group or factor comparisons
Stats_Pairwise Pairwise comparisons
Stats_KPSS Enabled KPSS and mixed-effects results
Stats_Effect_Sizes Effect-size estimates
Analysis_Coverage Analyses available for each video
Module_File_Index Per-video optional-analysis files
Module_Table_Index Consolidated optional-analysis tables
Figures_Index Exported figures
Assay_Figure_Index Figures recommended for the selected assay preset

An empty statistics or optional-analysis sheet means that the workbook structure was preserved but the batch did not meet the requirements for that analysis.

batch_session.json

This file stores the Batch Wizard queue and selected settings. Keep it with the results if you may need to resume, audit, or repeat the batch.

batch_results_status.json

This file records whether the batch workbook contains automatic results, finalized reviewed results, or results that need to be rebuilt after review changes.

Use the status message shown in the Batch Wizard rather than editing this file.

analysis_coverage_table.csv

This table is the quickest way to identify analyses that were unavailable or skipped for particular videos. Common reasons include:

  • missing ROI or object annotations
  • missing pose keypoints
  • a single-animal recording selected for a multi-animal metric
  • a single-class model selected for behavior-transition analysis
  • missing study-design labels for group statistics

Per-video inference files

Each video's inference folder may contain:

File What it contains
labels/<frame>.txt Detection or pose labels for individual frames
labels/labels.csv All available labels in one table
inference_metadata.json Video, model, frame-rate, and inference information
metrics.csv Per-frame movement measurements
metrics_summary_by_track.csv Per-animal movement summary
metrics_summary_by_frame.csv Per-frame animal count and movement summary
<video>_annotated.mp4 Optional inference video with overlays

Distances and positions in these files use pixel-based column names such as total_path_length_px. They are not physical distances unless the recording has been calibrated separately.

Per-video behavior files

The analytics folder always preserves the table structure for behavior bouts, even when no bouts meet the selected minimum duration.

File What it contains
<video>_detailed_bouts.csv One row per qualified behavior bout
<video>_summary.csv Bout counts and durations summarized by animal and behavior
<video>_bout_summary.xlsx Spreadsheet summary when qualifying bouts are available
<video>_analytics_dashboard.png Overview of behavior and ROI results
<video>_annotated.mp4 Optional analytics video with behavior, ROI, and object overlays
run_manifest.json Settings, source information, and file locations for this video's completed analysis

ROI files

ROI files are created only when arena ROIs are used.

File What it contains
<video>_roi_events.csv Entry and exit events
<video>_roi_overview.csv Concurrent ROI totals; nested ROIs can all receive occupancy
<video>_roi_per_track.csv Concurrent ROI totals for each animal
<video>_roi_dwell_events.csv Individual concurrent-membership visits
<video>_roi_exclusive_overview.csv Totals using one primary ROI at a time
<video>_roi_exclusive_per_track.csv Primary-ROI totals for each animal
<video>_roi_exclusive_dwell_events.csv Individual primary-ROI visits
<video>_roi_transitions.csv ROI-to-ROI transition totals
<video>_roi_transitions_per_track.csv ROI transitions for each animal
<video>_roi_behavior_summary.csv Behavior summarized within all active ROI memberships
<video>_roi_behavior_by_track.csv Concurrent ROI behavior summary for each animal
<video>_roi_exclusive_behavior_summary.csv Behavior summarized by primary ROI
<video>_roi_exclusive_behavior_by_track.csv Primary-ROI behavior summary for each animal

Use the exclusive files when you need categories that do not overlap. Use the concurrent files when nested or overlapping ROIs should each receive credit.

Object-interaction files

Object files are created only when object interaction is enabled and pose keypoints are available.

File What it contains
<video>_object_interactions_summary.csv Interaction totals by object
<video>_object_interactions_per_track.csv Interaction totals by animal and object
<video>_object_interactions_dwell_events.csv Individual qualified object visits
<video>_object_interactions_per_frame.csv Frame-level distance and interaction state
<video>_object_events.csv Object entry and exit events
<video>_object_approach_retreat_summary.csv Overall approach and retreat measurements
<video>_object_approach_retreat_by_track.csv Approach and retreat measurements by animal
<video>_object_approach_retreat_events.csv Individual approach and retreat events

The interaction distance is measured in original-video pixels from the selected keypoint to the nearest edge of the object ROI.

Review files

Review produces three kinds of records. Most users only need to know where to resume a review and where to find the latest exported tables. The detail below is useful when archiving a study or tracing how a reported result was produced.

1. Resumable review workspace

bout_review_workspace/
|-- IntegraPose_bout_review.sqlite3
`-- last_review_status.json

The workspace stores decisions, corrections, reviewer identities, and completion states. It allows the review to resume after the window closes.

Do not edit the database manually. Move it with the complete analytics folder.

2. Timestamped review exports

Every Export review snapshot action creates a new folder:

bout_review_exports/
`-- IntegraPose_bout_review_<timestamp>/
    |-- EXPORT_INDEX.md
    |-- review_export_manifest.json
    |-- Shared_Audit/
    |-- Behavior_Bouts/
    |-- ROI_Bouts/
    `-- Object_Interactions/

Only applicable categories are created.

Shared_Audit/ contains:

File What it contains
review_scope_status.csv Completed and provisional review scopes
review_audit_log.csv Reviewer actions and timestamps

Each applicable category contains Tables/ and Figures/.

Common category tables include:

File What it contains
original_predictions.csv Unchanged IntegraPose event predictions used to start review
review_decisions.csv All active and superseded review rows
corrected_bouts.csv Active accepted, modified, and manually added reference bouts
prediction_vs_review_scores.csv Event, frame, tIoU, agreement, and boundary metrics
bout_count_and_dwell_summary.csv Original-versus-reviewed counts and durations

Behavior review also includes:

File What it contains
behavior_correction_metrics.csv Per-class boundary, class, track, rejection, split, merge, and addition counts
behavior_overlap_review.csv Same-track overlaps and acknowledgement status
behavior_bout_transitions.csv Original-to-reviewed behavior changes
behavior_class_transition_matrix.csv Reclassification matrix
corrected_behavior_frames.csv Positive reviewed frames for every class and track
Per_Video_Bout_Tables/<video>_reviewed_behavior_bouts.csv Publication-ready reviewed behavior intervals

ROI and object review include corrected entry/exit markers and separate per-video dwell-event tables. Concurrent ROI and exclusive ROI-X intervals remain separate.

Figures summarize:

  • original and reviewed bout counts
  • original and reviewed dwell time
  • default tIoU or the advanced tIoU sweep
  • matched-event boundary error when available
  • concurrent and exclusive ROI counts when applicable

Scores and figures remain provisional wherever the corresponding scope has not been marked complete.

3. Preferred completed outputs

When a completed scope is exported, IntegraPose creates compatible reviewed tables under:

IntegraPose_bout_review_<timestamp>/
`-- IntegraPose_Authoritative/

The run record is then updated to identify those reviewed tables as the preferred source for later IntegraPose summaries. An incomplete scope does not replace the original automatic results.

For behavior review, every track must be complete. For ROI review, both the concurrent ROI and exclusive ROI-X scopes must be complete. Object interaction has its own completion state.

After reviewing a batch, choose Finalize Reviewed Results in the Batch Processing Wizard. This rebuilds the batch workbook and related summaries from the preferred reviewed outputs.

Older runs or legacy fallback tools may also contain flat files such as <video>_reviewed_bouts.csv or <video>_reviewed_roi_events.csv. Use the paths registered in run_manifest.json rather than choosing between similarly named files by hand.

Group statistics

The group_stats/ folder contains:

File What it contains
group_stats_overview.csv Overall tests across factor levels
group_pairwise_tests.csv Pairwise comparisons and adjusted p-values
group_effect_sizes.csv Effect sizes for the reported comparisons
group_kpss_mixed_effects.csv Enabled KPSS diagnostics and mixed-effects results

See Advanced Batch Statistics before interpreting these files.

Figures

Publication figure export can create both PNG and SVG versions. Depending on the selected export mode and available data, figures may include:

  • a batch dashboard
  • group-statistics overview
  • per-video summary profiles
  • path and speed traces
  • kinematic time series
  • ROI dwell profiles
  • object-interaction profiles
  • group distributions and time courses
  • omnibus significance heatmaps
  • pairwise effect-size plots
  • activity-budget, occupancy, transition, preference, latency, visit, and event-aligned summaries

Open figures/figure_manifest.csv to see every generated figure, its source table, and its location. The assay figure index provides a shorter list matched to the selected assay preset.

Optional-analysis tables

Optional analyses write their original per-video files inside the video's analytics folder. IntegraPose also gathers compatible files into module_tables/ for batch-level use.

  • module_file_index.csv lists the original per-video files.
  • module_table_index.csv lists consolidated tables and their row counts.

See Optional Analytics Reference for the purpose and requirements of each available analysis.

Files to keep together

For a complete, reproducible batch record, keep:

  • the entire chosen output folder
  • batch_session.json
  • each video's run_manifest.json
  • each reviewed video's bout_review_workspace/
  • all review exports used for reporting
  • the original videos or a stable archive location
  • the model and dataset YAML used for the run

Moving individual CSV files away from their manifests makes later review and Tab 7 handoff harder.

YOLO frame-level text files are required when Bout Analytics still needs to be run or repeated. They are not required merely to reopen a review whose analytics tables, manifest, and review video remain available.