CaltechAUTHORS
  A Caltech Library Service

A deep active learning system for species identification and counting in camera trap images

Norouzzadeh, Mohammad Sadegh and Morris, Dan and Beery, Sara and Joshi, Neel and Jojic, Nebojsa and Clune, Jeff (2020) A deep active learning system for species identification and counting in camera trap images. Methods in Ecology and Evolution . ISSN 2041-210X. (In Press) https://resolver.caltech.edu/CaltechAUTHORS:20201015-152731988

[img] PDF - Supplemental Material
See Usage Policy.

231Kb

Use this Persistent URL to link to this item: https://resolver.caltech.edu/CaltechAUTHORS:20201015-152731988

Abstract

1. A typical camera trap survey may produce millions of images that require slow, expensive manual review. Consequently, critical conservation questions may be answered too slowly to support decision‐making. Recent studies demonstrated the potential for computer vision to dramatically increase efficiency in image‐based biodiversity surveys; however, the literature has focused on projects with a large set of labeled training images, and hence many projects with a smaller set of labeled images cannot benefit from existing machine learning techniques. Furthermore, even sizable projects have struggled to adopt computer vision methods because classification models overfit to specific image backgrounds (i.e., camera locations). 2. In this paper, we combine the power of machine intelligence and human intelligence via a novel active learning system to minimize the manual work required to train a computer vision model. Furthermore, we utilize object detection models and transfer learning to prevent overfitting to camera locations. To our knowledge, this is the first work to apply an active learning approach to camera trap images. 3. Our proposed scheme can match state‐of‐the‐art accuracy on a 3.2 million image dataset with as few as 14,100 manual labels, which means decreasing manual labeling effort by over 99.5%. Our trained models are also less dependent on background pixels, since they operate only on cropped regions around animals. 4. The proposed active deep learning scheme can significantly reduce the manual labor required to extract information from camera trap images. Automation of information extraction will not only benefit existing camera trap projects, but can also catalyze the deployment of larger camera trap arrays.


Item Type:Article
Related URLs:
URLURL TypeDescription
https://doi.org/10.1111/2041-210x.13504DOIArticle
ORCID:
AuthorORCID
Norouzzadeh, Mohammad Sadegh0000-0002-2983-9374
Beery, Sara0000-0002-2544-1844
Additional Information:© 2020 British Ecological Society. Accepted manuscript online: 14 October 2020.
Subject Keywords:active learning; camera trap images; computer vision; deep learning; deep neural networks
Record Number:CaltechAUTHORS:20201015-152731988
Persistent URL:https://resolver.caltech.edu/CaltechAUTHORS:20201015-152731988
Official Citation:Sadegh Norouzzadeh, M., Morris, D., Beery, S., Joshi, N., Jojic, N. and Clune, J. (2020), A deep active learning system for species identification and counting in camera trap images. Methods in Ecology and Evolution. Accepted Author Manuscript. doi:10.1111/2041-210X.13504
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:106086
Collection:CaltechAUTHORS
Deposited By: George Porter
Deposited On:16 Oct 2020 19:55
Last Modified:16 Oct 2020 19:55

Repository Staff Only: item control page