Authors:

Jiawen Lyu

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

Video analytics has the potential to analyze animal behavior and identify individual cows. Retrieving all relevant video clips for each cow from unlabeled videos using cow recognition enables us to monitor individual cows. This research aims to develop a system that can effectively find all video clips for each cow from videos recorded in commercial settings. Original side-view and top-view videos, collected from January to April 2024, are taken from two cameras at a dairy farm. A deep-learning model is applied to the top-view videos to create a catalog (cattlog) storing each cow's coat pattern represented by bit vectors. The system compares each cow's coat pattern in the unlabeled top-view videos with each coat pattern in the cattlog to identify each cow. The system records the moments when a cow enters and leaves the camera view as the start and end times, respectively. Using this information, the system segments the unlabeled videos to generate all relevant video clips for each cow. We test the system for the accuracy of identification predictions and the proportion of specific cows of interest that are successfully found. The proposed system enables better monitoring of specific cows, facilitating analysis of health conditions and behaviors. Keywords: [no keywords provided]

Source:

Purdue University / 2024

Topics:

No topics listed

Co-authors:

Jiawen Lyu