You have unsaved work on - from before the page reloaded.
Drag a video here, or
Already tracked this one?
Working through several trials?
New here? - fetches a real sample clip and its saved analysis, no video of your own required.
You already have - video(s) saved from a previous folder. This selection has -.
The route the mouse took, colored dark to yellow as the trial goes on - click a point to seek the video there.
What does this mean?
Trajectory from trial start to the end of the recording, colored by 8 equal bands of elapsed time (viridis - perceptually uniform, colorblind-safe, and survives greyscale printing; banded rather than a smooth blend so neighboring time periods stay visually distinct). The bands are a fixed, even split of time, not a detected behavioral phase - a color change marks a clock boundary, not an event. Click a point to seek the video there. Failed-tracking frames break the line rather than bridging it - a gap is never silently interpolated.
How much time was spent where on the platform, not the route taken - click a cell to seek the video there.
What does this mean?
Occupancy over the platform, trial start to end of recording - how much time was spent where, not the route taken (see Path for that). Colored in the same 8 equal bins as Path, but here the bins are by visit density, not time. Click a cell to seek the video to the first frame that landed there.
One row per hole, one mark per visit, left to right in time - click a mark to seek the video there.
What does this mean?
One row per hole, target hole marked with a star (not color alone), trial start to end of recording left to right. Each mark is a real hole-visit, shaped by type (poke/entry/thigmotaxis) so meaning never depends on color alone. Click a mark, or focus the raster and use arrow keys plus Enter, to seek the video there - this tab, unlike Path and Heat map, is fully keyboard-operable.
Video & file metadata
- File
- -
- Codec
- -
- Resolution
- -
- Total frames
- -
- Keyframes
- -
- FPS (exact)
- -
- Duration
- -
- Duplicate-timestamp frames
- -
- Current frame
- -
- Frame timestamp
- -
Mouse tracking
Finds the mouse in each frame by comparing it against the empty-platform background, and marks a frame failed rather than guessing when it can't.
What does this mean?
Background subtraction against the median, gated to the confirmed platform circle and a plausible size range. The largest matching blob wins each frame. A frame with no valid blob is marked failed, never guessed.
What do these mean?
- Mask threshold
- How different from the empty-platform background a pixel has to be before it counts as mouse. Lower values pick up more of the animal but also shadows and lighting drift; higher values lose the animal in dimmer frames. The default of 30 tracked all three test clips despite very different lighting between the three rigs, which is the only evidence there is - no other value has been tested, and the 1-255 range is the arithmetic limit of an 8-bit image, not a tested range.
- Min blob area (px)
- Blobs smaller than this many pixels are rejected before the mouse is chosen. The default of 150 is already at the edge: test53's smallest observed mouse blob measures exactly 150, so partial views of the animal mid-poke are being clipped at the current setting. If tracking fails during hole investigations specifically, lowering this is the first thing to try. Raising it will lose more frames, not fewer.
- Max blob area (px)
- Blobs larger than this many pixels are rejected. The largest valid mouse blob observed across the three clips was 777 pixels, so the default of 1500 has roughly two-fold headroom; the fixed bolt this gate was added to exclude measured 3121, well above it. That gap is wide enough that this dial is unlikely to need changing, and no value other than 1500 has been tested.
- Tracked
- -
- Failed frames
- -
- Trial start (candidate)
- -
- Compute time
- -
A timeline of where tracking failed across the whole video - click anywhere to jump there.
What does this mean?
Where failures happened, not just how many - the whole video, start to end. Click anywhere to jump there.
Scrub the frame viewer above - green crosshair = tracked, red X = failed frame.
Save a portable .bmp.json file with everything except the video itself (which stays wherever it already lives) - open it later, on any computer, to resume without re-tracking.
Measures
Click a measure's name to see what it means and how it's calculated.
What does this mean?
Click a measure's name for its definition and how it's calculated. A row marked audited was changed by a human-review override above - re-click "Compute measures" after changing one to recalculate.
| Measure | Value |
|---|
Search strategy
Which search pattern this trial matches - spatial, serial, or random - or UNCLEAR when no single pattern fits.
What does this mean?
Spatial (direct) / serial / random. The definitions of the three strategies come from Gawel et al. 2019 (Table 1, citing O'Leary & Brown), including the "at most 1-2 holes adjacent to the target" limit that decides Direct - those are the literature's, not invented here. The 0-1 scores and the two bars that gate them are this project's own, and are described in the threshold notes below. A verdict requires exactly one categorical pattern to match; otherwise it's UNCLEAR and all three scores are shown for you to audit against the hole sequence, rather than forcing a guess.
What do these mean?
- Serial bar / Random bar
- A trial is called serial or random only if its respective score clears this bar. Unlike the strategy definitions themselves, these two numbers are not from the literature -
serialScoreandrandomScoreare this project's own 0-1 metrics (adjacency and direction consistency; platform-centre crossings), and the defaults of 0.6 and 0.5 are locally-set cutoffs on them. They were set from the shape of the literature's categories rather than calibrated, on the explicit reasoning that three test clips cannot calibrate a classifier. No value other than the defaults has been tested, and moving them is not a way to resolve an UNCLEAR verdict: a verdict also requires a categorical pattern match, so a score clearing its bar is necessary but not sufficient.
Export
One spreadsheet row for this trial, with every measure, the search-strategy verdict, and the thresholds that produced them.
What does this mean?
One tidy row for this trial: every measure above, the search-strategy verdict, the thresholds that produced them, and the tool version - opens directly in Excel, no legend needed. "Video file" is always included, so every exported row stays traceable on its own.
One spreadsheet row per hole visit, for looking at individual events instead of the trial total.
What does this mean?
Per-event detail: one row per hole visit (matches Merged Visits above, overrides included) - the "long form" alongside the trial summary's "short form."