Senior Design Projects
ECS193 A/B Winter & Spring 2023
AI-based animal recognition using trail camera images
| ********** | |
| Gary | |
| Natural Reserve System |
Project's details
| AI-based animal recognition using trail camera images | |
| The Stebbins Cold Canyon Reserve is a protected outdoor research laboratory and home to incredible amounts of wildlife, including deer, bears, mountain lions, bobcats, and foxes. To monitor these animals, we deploy trail cameras that collect color images and video. Many similar protected habitats employ this type of remote sensing, and the UC Natural Reserve System is endeavoring to standardize a process as part of a statewide effort to build a sentinel site network. But, trail cameras are notorious for collecting an abundance of images and the majority of which that are essentially “blanks” – no detectable wildlife. This creates a tremendous burden for researchers to streamline data collection and quality control. We hope to develop a workflow, with AI tools, to process trail camera data that can be expanded for multiple cameras and locations. | |
| The aim of this project is to create an AI-based workflow for images processing from trail cameras. Students should draw on the rich body of literature and established practices to assess what is currently happening in the field of image AI and what aspects they may want to incorporate into their project. Students will explore species level identification as an advanced step of the process and consider machine vs. crowd sourcing options. Students will need to consider how they might develop an AI to calibrate for different sites and species. As a part of the calibration and QC process, students would also need to determine best approaches to ascertaining probabilities with detection and accuracy metrics per identification. | |
| - Overview of current practices, state of image AI (literature review, summary); - Determine most feasible approach to development of an AI given the background; - An AI that: Assesses presence/absence of wildlife in images; An image-processing pipeline that is adaptable to other sites in California - Identifies species with some bounded CI or accuracy metric; - Calculation of accuracy metrics for identifications. |
|
| - Knowledge of cameras and imagery analysis; - Knowledge or experience with database APIs (retrieving data), pre-processing data, and data visualization tools - Basic understanding of statistics, probabilities, and accuracy |
|
| ********** | |
| 30-60 min every two weeks | |
| Open source project | |
| Attachment | N/A |
| Yes | |
| Team members | Brinda Puri Sanskriti Jain Zhantong Qiu Shuban Ranganath |
| Rex Liu | |
| N/A |