User Tools

Site Tools


9a.028.UL - Deep Learning Tool for Automatic Detection of Invasive Tegus from Camera Trap Photos

Project - Summary


  • Scientists collect more and more images in the field to study invasive animals in the wild
  • Going through all of these images to find those that contain animals takes resources
  • The goal is to automate the detection process
  • Deep learning based object detection
    • Transfer learning to enable use of small data sets
    • Object detection’s core is image classification
      • Inception-Residual architecture for better image classification performance
  • Build a web tool to collect more data

Project - Team

Team Member Role Email Phone Number Academic Site/IAB
Henry Chu PI (337) 482-0617 UL Lafayette
Keying Xu Co-PI (337) 482-0600 UL Lafayette
Rashida Hasan Student N/A UL Lafayette
Scott Wilson Project Mentor Office: (337) 266-8644 Cell: (337) 258-5557 USGS-WARC
Hardin Waddle Project Mentor N/A USGS-WARC
Funded By: USGS

Project - Novelty of Approach

See “Project - Summary”

Project - Deliverables

1 The primary product will be a user interface accessible via the internet to allow batch uploading of image files for analysis by established model algorithms, and fine-tuning of the method based on additional input data.
2 We will prepare a manuscript for journal submission describing the software and technique.
3 We will offer a webinar to the USGS Energy and Wildlife Community of Practice.

Project - Benefits to IAB

Object detection in images has many applications in other domains.

Project - Documents

projects/year9/9a.028.ul.txt · Last modified: 2021/06/02 16:29 by sally.johnson