Inputs and file layout¶

What each pipeline reads, and where it looks for it. Nothing here has to be renamed or moved: STRIDE writes only into its output folder.

Videos¶

.mp4, any letter case of the extension, in one folder or many. A folder is searched recursively; folders named waste are skipped (exclude_dirs).
Any file names. Names must be unique within a run: every table is keyed by the file name.
A name of the form Day<N>_<mouse>_Trial<T>.mp4 is recognised; day, mouse and trial are then filled from it. The mouse part may be any text (1042, M3-2, WT_1_1).
The frame rate is read from the video; fps_default (120) is used only when it cannot be read.

The stage-by-stage batch runner (run_all.py, stride analyze-decisions, stride run) requires the Day<N>_<mouse>_Trial<T> form and a reward table; stride tmaze and stride gait do not.

Pose files¶

One SLEAP predictions file per video, <video name>.slp (another ending with --pose-suffix / pose_suffix). The pipelines look next to the video first, then in the output folder's pose/ (files made by an earlier run), then in --pose-dir / pose_dir. One animal per video; the first instance of each frame is used.

A pose file remembers the path of the video it was made from. When that path cannot be read on the machine running the analysis, STRIDE uses the video on disk instead.

Keypoints STRIDE reads¶

The 15-node mouse skeleton of the original study:

snout, mouth, forepawR2, forepawR1, forepawL1, forepawL2, hindpawR2, hindpawR1, hindpawL2, hindpawL1, tailbase, tail1, tail2, tail3, tailtip

step	keypoints it reads by name
decisions (`stride tmaze`)	`snout`, `forepawL2`, `forepawR2`, `hindpawL2`, `hindpawR2`, `tailbase` (at least three must exist)
gait	`snout`, `tailbase`, `tailtip`, `hindpawL1`, `hindpawL2`, `hindpawR1`, `hindpawR2`

stride gait also accepts other skeletons: map their node names with keypoints:. It needs the snout, the tail base and at least one landmark per hind paw; the tail tip is optional (STRIDE Gait).

ROI files (T-maze only)¶

Seven polygons per video. stride tmaze looks in the output folder's roi/ (files made by an earlier run), then in --roi-dir / roi_dir, then next to the video, for these names, in this order: <video name>.preds.v2.best1.yml, <video name>.rois.yml, <video name>.yml. It writes them to the output folder's roi/ when it runs your ROI model.

image_file: /data/videos/clip_001.mp4
roi_count: 7
rois:
- id: 1
  name: arm_right          # arm_right, junction, arm_left, segment1 ... segment4
  type: polygon
  coordinates: [[x, y], [x, y], [x, y], [x, y]]

segment1 to segment4 tile the stem; segment4 is the one next to the junction and segment1 the far end. When no scale is given and no ChArUco board is in view, segment2 is the scale reference: the mean length of its shorter pair of opposite edges is taken as the corridor width (maze_width_cm, 10 cm).

What the ROI model must provide¶

A SLEAP top-down model with the 23-node maze skeleton: 16 corners and 7 centroids.

corners (16, all required)	centroids (7)
`arm_right.top_left`, `arm_right.top_right`	`arm_right.centroid`
`junction.top_left`, `junction.top_right`, `junction.bottom_left`, `junction.bottom_right`	`junction.centroid`
`arm_left.bottom_left`, `arm_left.bottom_right`	`arm_left.centroid`
`segment1.top_left`, `segment1.bottom_left` ... `segment4.top_left`, `segment4.bottom_left`	`segment1.centroid` ... `segment4.centroid`

How a frame is chosen and what happens when a corner is hidden: Pose and ROI models.

Metadata (optional)¶

`stride tmaze`¶

Without metadata the arm choice, the timing and the probe columns are written and the correctness columns stay empty. A metadata CSV (several may be given) has trial rows and/or mouse rows. Column names are case-insensitive.

rows	key columns	value columns
trial rows	`video` (file name, with or without extension) or `day` + `mouse` (+ `trial`)	`reward` (`L` / `R`); optional `day`, `mouse`, `trial`, `sex`, `genotype`
mouse rows	`mouse`	`sex` (or `Gender`), `genotype`

day + mouse rows are matched to videos named Day<N>_<mouse>_Trial<T> (the day may be written Day1, DAY1 or 1); for any other file name use the video column. A video without a matching row keeps empty correctness; the run says how many.

day,mouse,reward
Day1,101,R
Day2,101,L

Mouse,Gender,Genotype
101,F,WT
102,M,WT

`stride gait`¶

A CSV with a video column holding the file name and any other columns. An animal id (mouse_id, mouse, animal, animal_id, subject or id) fills mouse_id; sex / gender fills sex; genotype fills genotype; day fills day. Every other column, and the id column under its own name, is added at the end of the gait tables.

video,animal,sex,group,weight_g
runway A1.mp4,A1,F,control,21.5

Batch runner (`run_all.py`)¶

meta_trials_csv with the columns day,mouse,reward and metadata_csv with one row per mouse (Mouse,Gender,Genotype; the names are mapped in the batch's column_map). In these entry points a video whose (day, mouse) has no reward row is written as not correct (correct_bin = 0) and a warning names it; stride tmaze leaves the correctness empty instead.

Scale¶

See Scale calibration. In short: a value you give, else the ChArUco board in view, else (T-maze only) the ROI corridor width.