Skip to content

4. Retrieving Detection Statistics

Learning Goal: Access the statistics stored in fmdt-detect's log files.

In order to enable the creation of log files, we must specify a directory using the log_path parameter.

res = fmdt.detect(vid_in_path="demo.mp4", log_path="log")

During detection, fmdt-detect will store important statistics for each frame. The 5th frame, for example, will have its statistics stored in the file ./log/00004.txt.

We've actually already done the heavy lifting and stored a few relevant statistics in our res. To extract them to a pandas.DataFrame, we call the to_dataframe() function.

>>> print(res)
fmdt.res.DetectionResult with args digest: 137719b27ea97a86
objects in trk_list: 38 meteor(s), 0 star(s), 0 noise
     nroi  nassoc  mean_err  std_dev
0      45       0    0.0000   0.0000
1      63      17    0.1669   0.1450
2      67      19    0.2917   0.3966
3      82      24    0.2733   0.2614
4      35      18    0.1626   0.2126
..    ...     ...       ...      ...
251    52      18    0.3126   0.2950
252    59      22    0.9310   1.2554
253    65      17    0.1500   0.1209
254    50      17    0.1123   0.1086
255    66      19    0.3718   0.2947

[256 rows x 4 columns]

Where, for each frame:

  • nroi is the number of regions of interest,
  • nassoc is the number of associated regions between frame t - 1 and frame t,
  • mean_err and std_dev are motion estimation statistics explained here, also between frame t - 1 and frame t.

IMPORTANT: If the argument log_path is not provided then trying to call print(res) will not show the motion statistics.