Datasets and code (Part2) associated with "From video to behaviour: an LSTM-based approach for automated nest behaviour recognition in the wild"

This Zenodo record contains the second part off the materials needed to replicate the study “From video to behaviour: an LSTM-based approach for automated nest behaviour in the wild", found here https://www.biorxiv.org/content/10.1101/2025.05.27.656097v5.article-info

This archive represents Part 2 of a two-part material deposited on Zenodo, divided due to file size limitations.

This second part contains:

i) All scripts required for modelling and model evaluation, independent of bird species (present in both parts)
ii) A comprehensive README file providing detailed instructions to reproduce all the modelling and analyses presented in the study (present in both parts)

iii) The datasets required for LSTM modelling of aggression and building detection in sociable weavers (Philetairus socius) 

iii) The datasets required to assess the effects of training dataset size, hard negatives proportion, and the adoption of YOLO  on model performance and deployment when automating nest activity detection in sociable weavers

iv) The datasets required for LSTM-based nest activity and sanitation recognition in the blue tit (Cyanistes caeruleus)

v) The dataset required for LSTM-based nest activity recognition in the great tit (Parus major)

vi) The LSTM-based models and CSV dataframes provide guidance for the two behavioural automation tasks in the blue tits: nest activity and sanitation

vii) The LSTM-based model and CSV dataframes provide guidance for the nest activity automation in the great tits