FiNTA analysis export
=====================

Job:       fdcf9293-ca43-4b6e-8a4b-ec4e20c1230c
Image:     Franzi_CPD_012.tif
Status:    Succeeded
Finished:  2026-08-15T15:04:13.156Z
Runtime:   2 min 28 s

Contents
--------
report.pdf             Readable report: parameters, overlays, statistics,
                       histograms.
settings.set           The FiNTA settings file this run used, verbatim.
params.json            The structured parameter snapshot the settings file was
                       rendered from.
input/                 The image that was traced.
images/                The overlays FiNTA drew.
video/                 The animate_tracing recording — the raw .avi, a
                       browser-playable .mp4 transcode where available, and
                       its poster frame.
lines/                 draw_lines' per-fibre highlight images, one folder per
                       command.
raw/                   Raw tracing/loop geometry (save_tracing_data,
                       save_loop_data) — node/edge or polygon data with no
                       fixed shape; not parsed, download-only, can run to tens
                       of MB.
data/                  FiNTA's raw output files: one number per line for a
                       distribution, three whitespace-separated columns per
                       line for persistence_data — fibre index, segment
                       separation, direction correlation. The fibre index is
                       also the row number in that command's line-lengths
                       file, which is what makes a per-fibre persistence
                       length recoverable.
csv/summary.csv        FiNTA's own summary values.
csv/distributions.csv  One row per distribution: n, min, max, mean, median,
                       std. dev., total. Orientations leave those columns
                       empty — their angles wrap, so an arithmetic mean of
                       them is wrong rather than approximate — and carry
                       circular statistics in the axial_* columns instead.
csv/histograms.csv     One row per histogram bin.
csv/point_clouds.csv   One row per (x, y) pair of a thinned point cloud: the
                       raw correlation decay, and one row per fibre of fibre
                       length against that fibre's own fitted persistence
                       length (mean_y is the mean over fibres, which differs
                       from FiNTA's pooled network-wide fit).
metrics.json           The same numbers as the CSVs, nested.
finta-stdout.log       What FiNTA printed while running.

Units
-----
All lengths are in nm and all areas in nm², from the scale bar measured when
this job was created.
FiNTA applies the scale factor to its output but records the unit nowhere in
the files themselves — the unit above is carried by this application, not by
the .dat files in data/.

A note on the total mesh area
-----------------------------
FiNTA's own "total area of all loops" in csv/summary.csv disagrees with the
sum of its per-loop area file (the `total` column in csv/distributions.csv).
The summary value applies the scale factor twice and halves the shoelace sum
twice; the per-loop file is scaled exactly once. The distributions total is
the correct one. Both are reported unchanged — FiNTA itself is not patched.
