dd4wh / monitor

Auto-ID of bird calls, Pygmy Owl as an example

Sperlingskauz - Projekt 2020

Covid-19 prevented me from teaching my students in practical zoological exercises and in ornithological field excursions this spring. In March 2020 in Germany, leaving the house was only possible for doing sports and buying food.

So, I thought: lets do some sports! I left my house for sports, entered the huge mixed forest behind our house and put out twelve time-scheduled audio recorders for about one week.

The aim was to detect Eurasian Pygmy Owl which I knew was breeding in small numbers in the forest, but is known to be quite challenging to detect.

In the consequence, I gave 40,000 minutes of WAV audio recordings to my students (67 participants) and told them to search for Pygmy Owl within these recordings and write a report on this, which would make up their exam in my course this semester. At the same time, I used a specific kind of easy-to-use machine learning algorithm monitoR (Katz et al. 2016) to automatically detect the species in the recordings.

Additionally, a group of regional specialised ornithologists also made a synchroneous survey in the same sites on March, 25th with 24 ornithologists on-site . . .

so the competitive challenge could start!

And here are the results!

Site	Students	Auto-ID	Specialists
1	x	x	-
2	-	-	-
3	x	x	-
4	x	x	-
5	x	x	x
5b	x	x	x
6	-	-	x
7	x	-	?
8	?	-	x
9	-	-	-
10	-	?	x
11	-	-	-
12	-	x	x

The three methods seem to be equally suited to discover Pygmy Owl!

BUT: they are also complementary! So, I recommend a combination of ornithological work in the field with Auto-ID of recordings from the field.

Time-scheduled recordings of animal vocals with subsequent automatic classification can deliver a nice piece of additional information and can largely save time and money for nature conservation work!

Also, it can help to keep students doing valuable stuff in Covid-19 situations :-).

This github repository contains the code used for doing the auto-ID part of this fun and educational project.

HAVE A LOOK AT THE DETAILED INFORMATION IN THE WIKI ! [https://github.com/DD4WH/monitoR/wiki]