discussion / Acoustics  / 29 September 2023

Peruvian Bird Audio Dataset Questions

Hello everyone! 

I am part of a team on developing CNNs to detect species in the Peruvian Amazon from soundscapes. We are having trouble finding a ground truth dataset for the region to test our model against. We have been trying to use “A collection of fully-annotated soundscape recordings from the Southwestern Amazon Basin” (https://zenodo.org/record/7079124#.Y7iis-xudhE) which seems to be the best publicly available strongly labeled ground truth from the region. However, we think there are a few errors in the labeling leading to our models having poorer results than they might. For instance, in the screenshot below, there seems to be an unlabeled species between 34:38 and 34:30 from 1 to 4 hz (and repeated again at 34:35 to 34:38) that is labeled in annotations 15616 and 15622. We have noticed a handful of what we think are these errors in the dataset, but we don’t necessarily have the expertise to know for sure. Has anyone used this data in training/testing and if so what are your thoughts on this data?

Thank you for your time!

Sean Perry

Hi Sean!

Wanted to just mention that Arbimon, Rainforest Connection's ecoacoustic platform, has a number of projects in Peru (here, here, & others, if you search by keyword 'peru'). We have some existing CNNs for that region (mostly from Ecuador & Brazil, but there is likely species overlap). Do feel free to DM me here or email me ([email protected]) and I'm happy to talk about collaborating! 



Hi Sean, sounds like an excellent project. Definitely talk with the Arbimon folks! As @carlybatist said, I am working with birds in the Peruvian Amazon, but in Loreto. Definitely would like to hear more about your project. As for this labelling issue here, definitely looks like an error. I have not used this dataset, now I'm curious.  To be honest, some questionable labelling in that file in general. Bounding boxes can be very subjective...