Elemantra: An End-to-End Automated Framework Empowered with AI and IoT for Tackling Human-Elephant Conflict in Elephant-Range Countries
By Nuwan Bandara and Dilshan Bandara
If you find our work, this repository, or novel algorithms or annotated datasets useful, please consider giving a star ⭐ and citing our paper.
@misc{bandara2023elemantra,
title={Elemantra: An End-to-End Automated Framework Empowered with AI and IoT for Tackling Human-Elephant Conflict in Elephant-Range Countries},
author={Nuwan Sriyantha Bandara and Dilshan Pramudith Bandara},
year={2023},
eprint={2310.15012},
archivePrefix={arXiv},
primaryClass={eess.SY}
}
-
(Nov 1, 2023)
- Infrared Elephant Image Annotated Dataset ('EleThermal' dataset) is released.
-
(Oct 24, 2023)
- Base Python codes for the novel algorithms in Elemantra are released.
Refer requirements.txt
for the required essential packages.
pip install -r requirements.txt
Configure the datapaths in config.yaml for each task (i.e. seismic-based elephant detection, modified bee sound-based elephant repelling and infrared signature-based elephant detection) as in the folder structure of this repository after adding more data points, especially for seismic-based elephant detection and modified bee sound-based elephant repelling, as required.
Note: The sample data for bee sound processing and seismic signal processing are in sample_bee_sounds (retrieved from here) and sample_seismic_data folders (retrieved from here) respectively. The annotated dataset for infrared image-based elephant detection is in the dataset: elephant_detection (the respective images are collected from here and annotated by this project). Therefore, if you use, by any means, data and/or proposed algorithms, please cite the above sources accordingly, along with our work.
Execute the tasks
# execute entire pipeline with default task: seismic-based elephant detection
python3 main.py
Other forms of execution
# execute entire pipeline for the task: modified bee sound-based elephant repelling. Note that the spectrogram correlation calculations usually takes a considerable time to execute
python3 main.py --task=bee-modify
Elephant detection instances in test infrared set using tflite model
with high confidence | one retrieved from here | one which failed to detect multiple elephants |
---|---|---|
Elephant detection instance using seismic recordings
frequency distribution | detection in STFT plot | detection using proposed method |
---|---|---|
STFT spectrogram plots of one:
original bee sound segment of 4𝑠 | corresponding modified bee sounds through randomly changing the FR | corresponding modified bee sounds through randomly adding an overlay of pink noise |
---|---|---|
Authors would like to extend their gratitude to Dr. Mukunthan Tharmakulasingam from the University of Surrey, United Kingdom, Dr. Chamira Edussooriya, Ms. Anjalie Kalansooriya and Mr. Derek Nanayakkara from the University of Moratuwa, Sri Lanka for their valuable suggestions.