Phase 1 results

Dear Deep Sea Explorers,

We would like to thank you from the deepest of our heart - and the sea! - for the work you have already carried out on our project. We sincerely appreciate the help provided by each one of you!

We have been analysing the results of the classifications you have performed so far.

Bioluminescence workflow:

Thanks to you, we can now say that these two events are one peak-events with a 90% and 96% certainty (i.e., this is how often they were classified in this category by the participants!).

While this one is very likely (100% of your classifications) a multiple-peak one

There were some events however were do not seem to agree.

This one for example could be a 2 or multiple-peak event (55%-42% of your classification, respectively).

And this one could be a one-peak (33%), a multiple-peak (29%), or a weird (32%) event.

Here is how were classified 50 of the events you have worked with, where you determined they were more likely one-peak or multiple-peak events, respectively:

.

Bioacoustic workflow:

Similarly, some events were classified with a very high probability in one category.

This one is likely (84%) a sperm whale:

While this one is probably a short-finned pilot whale

This event however could be a short-finned pilot whale (28%), a sperm whale (35%), or pure noise (24%).

Here is how were classified 50 of the events you have worked with:

What does it mean?

As the goal of the project is to obtain a set of very well classified objects, we have prepared a phase 2 for the project and we need your help! The goal is to provide new subjects so we can increase the number of events that are classified with a high probability in only one category. We will then use these events to train a machine learning algorithm to recognize all the future events KM3NeT will record during its (hopefully very long) lifetime!

Phase 2 has many more events than phase 1, as we want to significantly increase our statistics.
Some of the events of Phase 2 might appear a bit strange, like these ones for example

In these events, the baseline might be lower than previously, the peaks less well defined, or strange data points forming another layer.

This is absolutely normal: remember we work with sensors immersed more than 2000 m deep in the sea and we are looking at living organisms – large variations in the signal are expected. This is exactly why we need you to decide what this event could be as a machine would not be able to deal with so large fluctuations in the patterns!

Are you ready for Phase 2 of the Deep Sea Explorer mission? 3, 2, 1 … Dive!

Sincerely,
The KM3NeT-REINFORCE team.