Skip to content

1. Your first meteor

Learning outcome: Learn how to launch fmdt-detect using fmdt

Prerequisites

In order to follow this tutorial you need to have

  • fmdt-python installed (see installation if not),
  • fmdt-detect installed (follow this link if not),
  • fmdt-detect on your system's PATH.

Start

In this tutorial we are going to learn how to detect meteors using the python module fmdt. We'll start off by retrieving a sample video, saving it as demo.mp4.

Let's open up a new Python script and jot down the following snippet.

hello_fmdt.py
""" A sample program to call `fmdt-detect` using fmdt-python"""
import fmdt

vid     = 'demo.mp4'
trk     = 'trk.txt'

fmdt.detect(vid_in_path=vid, trk_path=trk)

The last line executes fmdt-detect with the equivalent bash command:

fmdt-detect --vid-in-path demo.mp4 > trk.txt

Open up a terminal and lauch this script (or type in the previous statements in an interactive session) 

python3 first_detection.py

The function fmdt.detect uses the same parameters as the command-line utility fmdt-detect, the only difference being that flags like have their dashes - transformed into underscores _ since Python interprets dashes as minus signs. As such, the usual flag --vid-in-path is encoded as the python argument vid_in_path.

For a complete list of parameters, see API.

Visualizing

Before the visualization we need to generate intermediate files:

  • the log files (contains all the characteristics for each Regions of Interest (RoI) depending on the frame number),
  • a file that contains the correspondence between the track ids and the RoIs ids,
  • a file that contains the bounding boxes for each frames.

For this, we need to run fmdt.detect and to output 1) log files and 2) the "tracks to RoIs" file.

hello_visu_fmdt.py
""" A sample program to call `fmdt-detect` using fmdt-python"""
import fmdt

vid     = 'demo.mp4'
trk     = 'trk.txt'
trk2roi = 'trk2roi.txt'
log     = 'detect_log'

fmdt.detect(vid_in_path=vid, trk_path=trk, trk_roi_path=trk2roi, log_path=log)

Once it is done, we need to generate the bounding boxes with an intermediate program: fmdt.log_parser.

bb = 'bb.txt'
fmdt.log_parser(trk_roi_path=trk2roi, log_path=log, trk_bb_path=bb)

Then we can visualize the tracked meteors with a call to fmdt.visu.

fmdt.visu(vid_in_path=vid, vid_out_path="demo_v.mp4", trk_path=trk, trk_bb_path=bb)

In order to visualize, we need to tell fmdt where we stored the tracking list (trk_path) as well as the bounding boxes (trk_bb_path).

Conclusion

Congrats! You've detected your first meteor! 38 of them, if everything went as planned. Go ahead and follow the next tutorial to find out how to inspect the tracked objects.