Skip to content

README.md

Latest commit

 

History

History
40 lines (34 loc) · 2.76 KB

README.md

File metadata and controls

40 lines (34 loc) · 2.76 KB
Raw

Data

Datasets are loaded at runtime by each notebook (downloaded from this repo, or generated as a stand-in) so the repo stays Colab-friendly. Real data small enough to ship lives here in-repo.

What ships here

  • transition_data.mat (1.9 MB) — the real 59-fly transition dataset for 02_transitions_and_hierarchy.ipynb (defaults to USE_SYNTHETIC_DATA = False).
  • rat_data/ (~7 MB) — Ugne Klibaite's rats for 03_rat_individual_behavior.ipynb: amph.npz (6 rats × 3 days of precomputed individual-map (cz_action) embeddings + fine/coarse watershed labels; days 2–3 baseline, day 4 amphetamine; all 18 share one map space) and rat_keypoints_session1.npz (a 23-joint 3-D s-DANNCE keypoint clip + skeleton + one lone session's behavioral map, for the "look at the raw data" intro). Regenerate with python tools/make_rat_data.py; source is Dropbox/LE_CONTROL_AMPH/ (the _L lone sessions — same cohort as the social data below).
  • rat_data/rat_social.npz + rat_data/rat_social_keypoints.npz (~20 MB) — real Long-Evans dyads for 04_rat_social_behavior.ipynb: 45 dyads (24 control + 21 amph) + 30 lone sessions, each with the precomputed individual map (cz_action + coarse labels) and social/joint map (sz_joint + coarse labels) and partner labels (rat_social.npz), plus 23-joint 3-D keypoints for 2 example dyads + a viz clip (rat_social_keypoints.npz), down-sampled to 5 Hz. isamph: 0 = control, 1 = amphetamine, 2 = saline cage-mate of an amph rat. Regenerate with python tools/make_rat_social_data.py; source is Dropbox/LE_CONTROL_AMPH/.
  • optogenetic_data/*.npz (7 files, ~4 MB each) — all seven real driver lines from the Cande et al. 2018 example data (ss02635, ss02617_0226, ss01049, ss01540, ss01597_1v_1022, ss01602, ss02393_1v_1009), for 06_optogenetics.ipynb (defaults to the real data; the notebook's §9 is a menu of all seven). Each file is one 12-fly filming session: cameras 1–6 are controls (no all-trans-retinal), 7–12 are experimentals. Stored as the 2-D map embedding (out-of-hull points replaced by zGuesses, as in the MATLAB loadZValuesAndLEDs.m) plus a per-frame LED on/off trace, down-sampled from 100 Hz to 25 Hz as float32. Regenerate all seven with python tools/make_optogenetic_data.py; the source is in Dropbox/.../Cajal_behavior/Fly_Optogenetic_Analysis/example_data/.

07_bring_your_own_data.ipynb is the only notebook without bundled data — it's a template you point at your own tracking.

Still to wire up

All Act-2 datasets are now in the repo. 01_build_a_behavioral_map.ipynb (fly LEAP) and 05_slow_modes.ipynb (cached .npz/.pkl) download their data automatically by cloning motionmapperpy and slowmode — no action needed.