🗂️ Example Datasets

Videos, pose-estimation models, trained classifiers and training data shared by the SimBA team, grouped by experimental set-up. Use them to follow the tutorials, test SimBA on real data, or as a starting point for your own classifiers.

Looking for something specific? The data inventory lists every file the SimBA team has shared (on OSF, Google Drive and in the GitHub repository), with a search box, what is inside each archive, body parts, classifier settings and feature columns.

⬇ Downloading Large folders are split into zip volumes (.zip.001, .zip.002, …).
  1. Download all volumes in the folder into one directory.
  2. Extract only the first volume (.zip.001) with 7-Zip; it reads the others automatically.
⚠ Classifier compatibility The shared classifier files (.sav) were saved with an older scikit-learn. They load in SimBA’s Python 3.6 install (scikit-learn 0.22) but not on Python 3.9+ installs (scikit-learn 1.4). On newer installs, retrain from the included training data instead (Scenario 3).
🐭

Mouse resident-intruder

Mouse2 animalsVideosPose modelsClassifiers

A resident mouse and an intruder in the home cage, filmed from above. Videos come in RGB, CLAHE-enhanced and greyscale versions, with matching DeepLabCut models for white-vs-black mouse pairs.

Use with: Scenario 2 (run classifiers on new data)
🐀

Rat resident-intruder

Rat2 animalsVideosPose modelClassifiersTraining data

A resident rat and an intruder in the home cage, filmed in RGB, with a DeepLabCut model, seven trained classifiers and the annotated training data behind them.

Use with: Scenario 2 (run classifiers) · Scenario 3 (retrain or improve)
🎥

CRIM13 (Caltech Resident-Intruder Mouse)

Mouse2 animalsRaw videos: CaltechPose modelClassifiersTraining data

237 pairs of synchronised top- and side-view videos (88 hours, 8 million frames) annotated for 13 social behaviours, from Caltech. SimBA adds a DeepLabCut model, 11 trained classifiers and training data, so you can go from raw video to classified behaviour with public files only.

Cite: the CRIM13 videos are by Burgos-Artizzu, Dollár, Lin, Anderson & Perona — please cite their dataset (doi:10.22002/D1.1892).
Use with: Scenario 2 · Scenario 3
🍼

AMBER: maternal behaviour

RodentDam & pupsExample videoClassifiersSHAP

Classifiers from the AMBER pipeline for scoring maternal care: active and passive nursing, licking/grooming, nest attendance, dam eating and drinking, and self-directed grooming. Includes an example video and SHAP explanations.

Cite: Lapp, H. E., Salazar, M. G. & Champagne, F. A. (2023). Automated maternal behavior during early life in rodents (AMBER) pipeline. Scientific Reports (doi:10.1038/s41598-023-45495-4).
🟫

Mouse open field

Mouse1 animalVideosPose model

A single C57BL/6J mouse in an open-field arena, filmed in RGB, with a DeepLabCut model. A good starting point for ROI and movement analyses.

🔘

Mouse operant chamber

MouseVideosPose model

A C57BL/6J mouse in an operant chamber during social self-administration, filmed in RGB, with a DeepLabCut model.

All files are hosted on the SimBA OSF repository unless marked otherwise. Pose-estimation models for other tools (YOLO, DeepPoseKit, Mask R-CNN) and their labelled images are in the tracking models project.