User Tools

Site Tools


6a.060.UL - Detecting and Identifying Wildlife Animals From Images Using Deep Learning

Project - Summary

Wildlife detection, recognition, and tracking individual species are a very important aspect of ecological and environmental studies. Scientists are increasingly using digital technologies like infrared cameras that collect data to track and inventory these species; however, several of these efforts to identify and classify the species are done manually. Manual detection, identification, and recognition of these species become time consuming, expensive and sometimes unfeasible due to lack of human resources. Researchers at CVDI have expertise in computer vision and have done prior work in applying machine learning techniques for ecological informatics [1] [2]. The overall goal of this research program is to develop big data based pattern recognition techniques to solve computer vision problems associated with detection, recognition and tracking of individual species across multiple domains. We experimented with using two families of deep learning algorithms for object detection to locate Tegu lizards in digital images captured by motion-sensed cameras and achieved 75% detection rate with 0 false positive rate. We also experimented with preprocessing tasks necessary for identifying manatees in human-captured digital images.

Project - Team

Team Member Role Email Phone Number Academic Site/IAB
Henry Chu PI Not available (337) 482-0617 UL Lafayette
Keying Xu Co-PI (337) 482-0600 UL Lafayette
Scott Wilson Project Mentor Office: (337) 266-8644 Cell: (337) 258-5557 USGS Wetland & Aquatic Research Center

Project - Deliverables

1 Engineered features for fraud data set, with implementation on AWS
2 Definition and implementation of adversary and defense strategies
3 Analysis of ROCs under attack-defense combinations
4 Conclusions and final report

Project - Presentation Video

Project - Documents

projects/year6/6a.060.ul.txt · Last modified: 2019/08/14 16:32 by sally.johnson