/smart_camera_trap_research

Primary LanguageC++

Research on smart camera traps

Citing

Please cite our paper if you find this repository useful.

@article{VELASCOMONTERO2024102815,
  title = {Reliable and efficient integration of AI into camera traps for smart wildlife monitoring based on continual learning},
  journal = {Ecological Informatics},
  volume = {83},
  pages = {102815},
  year = {2024},
  issn = {1574-9541},
  doi = {https://doi.org/10.1016/j.ecoinf.2024.102815},
  url = {https://www.sciencedirect.com/science/article/pii/S1574954124003571},
  author = {Delia Velasco-Montero and Jorge Fernández-Berni and Ricardo Carmona-Galán and Ariadna Sanglas and Francisco Palomares},
}

https://www.sciencedirect.com/science/article/pii/S1574954124003571

Hardware

Compatible with Raspberry Pi devices. Tested, in particular, on Raspberry Pi 3B, Raspberry Pi 4B, and Raspberry Pi Zero 2W.

Installation

  1. Download this repository:
git clone https://www.github.com/DVM000/smart_camera_trap_research.git
  • Installation scripts assume that code is located at /home/pi/smart_camera_trap_research/
  • You must give execution permission to all scripts.
  1. Install dependences (it will take a long while):
cd scripts
sudo ./install-dependences.sh
cd ..
  1. Build the application. It will generate the executable file in /apps/APP
./scripts/COMPILE_APP.sh

Enable camera using raspi-config.

Note: tested on Rasbian GNU/Linux 10 (buster) kernel Linux 6.1.61-v8+ aarch64 32 bits with OpenCV version 4.4.0 and TensorFlow 2.1.0 (Python 3.7)

Configuration

Run this script to pre-configure the applicaction:

./scripts/settings.sh

You can use parameters by default or you can edit configuration.ini according to required application parameters.

You can use parameters by default or you can edit training_config.ini for configuration of on-device training parameters.

Usage

You can direcly run the application with:

sudo ./scripts/service_app/Launch_service.sh

Or you can also establish it as a system service for executing on boot:

sudo cp ./scripts/service_app/camera-trap.service /lib/systemd/system/  
sudo systemctl enable camera-trap.service
sudo reboot

Outputs

  • Application log will be ./OUTPUT/App_log_<date> . This log contains chronologically-ordered information about camera status, PIR activation, animal detection, re-training status, night detection, etc (including timestamps);
    as well as possible source or errors encountered during execution.
  • Classified frames will be located into ./OUTPUT/<classname> folders.
  • Periodically collected frames for on-device training are stored into ./DATASET/FRAMES/.
  • Training image datasets will be built into ./DATASET/img_data_XXX folders.

Advanced operation and hardware instructions

Take a look at here.