Skip to content

This repo contains code+pre-trained models for extracting information from camera-trap images. The pre-trained models have been trained on the Snapshot Serengeti dataset.

License

Notifications You must be signed in to change notification settings

sirogosig/CNN_camera_trap_images_Noroussadeh

 
 

Repository files navigation

deep_learning_for_camera_trap_images

This repository contains the code used for the following paper:

Automatically identifying, counting, and describing wild animals in camera-trap images with deep learning

Authors: Mohammad Sadegh Norouzzadeh, Anh Nguyen, Margaret Kosmala, Ali Swanson, Meredith Palmer, Craig Packer, Jeff Clune

If you use this code in an academic article, please cite the following paper:

	@article {Norouzzadeh201719367,
		author = {Norouzzadeh, Mohammad Sadegh and Nguyen, Anh and Kosmala, Margaret and Swanson, Alexandra and Palmer, Meredith S. and Packer, Craig and Clune, Jeff},
		title = {Automatically identifying, counting, and describing wild animals in camera-trap images with deep learning},
		year = {2018},
		doi = {10.1073/pnas.1719367115},
		publisher = {National Academy of Sciences},
		issn = {0027-8424},
		URL = {http://www.pnas.org/content/early/2018/06/04/1719367115},
		eprint = {http://www.pnas.org/content/early/2018/06/04/1719367115.full.pdf},
		journal = {Proceedings of the National Academy of Sciences}
	}

Most of the code in this repository is taken from here

This repository has four independent parts:

1- The code used for Task I: Detecting Images That Contain Animals (phase1 folder)

2- The code used for Task II,III, and IV: identifying, counting, and describing animals in images (phase 2 folder)

3- The code used for Task II only, (all the transfer learning experiments for Task II used this part of the repo) (phase2_recognition_only folder)

4- resize.py is used for resizing the input images for all the other parts

For more information on how to use this repo please refer to the base repo at this link

1. Requirements

Requirements

To use this code, you will need to install the following:

  • Python 2.7
  • Tenorflow
  • NumPy
  • SciPy
  • MatPlot Lib

2. Running

Pre-trained models could be found at the following links:

  • Phase 1 (VGG architecture):

https://drive.google.com/file/d/1Y-aDWNMfvgYUb-u-_cqzibZ6ePFOOLGj/view?usp=sharing

  • Phase 2 (ResNet-152 architecture):

https://drive.google.com/file/d/1KTV9dmqkv0xrheIOEkPXbqeg36_rXJ_E/view?usp=sharing

  • Phase 2 recognition only (ResNet-152 architecture):

https://drive.google.com/file/d/1cAcnyBTO5JeB2zSaEoGBWf0Jd-jAnguS/view?usp=sharing

3. Licenses

This code is licensed under MIT License.

4. Questions?

For questions/suggestions, feel free to email, tweet to @arashnorouzzade or create a github issue.

About

This repo contains code+pre-trained models for extracting information from camera-trap images. The pre-trained models have been trained on the Snapshot Serengeti dataset.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 78.8%
  • Jupyter Notebook 21.2%