Redefining Lameness Assessment Using Crowd-Sourced Data

This repository contains the data and code for our project: Which Cow is Most Lame? Redefining Lameness Assessment Using Crowd-Sourced Data.

Example Videos in Lameness Hierarchy

Repository Structure

Here's a brief overview of the repository's structure. The prefix number in each folder name indicates the sequence of analysis. Each folder contains code and results in the sub-analysis:

• 00-easy_hard_question_cutoff: Contains machine learning models and output used in analyzing a old data set from phase 1 of this project, in order to distinguish video pairs with clearly distinguishable (easy) and hard to distinguish (hard) lameness differences between the 2 cows.
• 01-video_select_compress: Dataset that contains selected cow videos and code for video compression.
• 02-generate_30cow_GS_label_html_experts: Code and data for generating HTML files to score 30 cows based on 5 level locomotion scoring system.
• 03-30cow_GS_label_expert_response: 5 experts' (3 rounds per experts) answer regarding 30 cows' gait score.
• 04-generate_54HIT_html_experts: Code and data used in generating HTML to ask lameness experts to compare every 2 cows in the 30 cow group (435 pairs of comparisons) for pairwise lameness assessment.
• 05-Amazon_MTurk_expert_response_30cow_pairwise: 4 experts' responses to 435 pairs of lameness comparisons on Amazon MTurk.
• 06-generate_54HIT_html_click_worker: Code and data used in generating HTML to ask click workers from Amazon MTurk to compare every 2 cows in the 30 cow group for pairwise lameness assessment.
• 07-Amazon_MTurk_click_worker_response_30cow_pairwise: 20 click workers' responses to 435 pairs of lameness comparisons on Amazon MTurk.
• 08-Lameness_rank_eloSteepness: Lameness rank generated based on pairwise lameness assessment using EloSteepness.
• 09-Lameness_rank_merge_sort: Lameness rank generated based on pairwise lameness assessment using merge sort.
• 10-Lameness_rank_borda_counting: Lameness rank generated based on pairwise lameness assessment using borda counting.

Thank you for your interest in our project. We hope you find the data and code insightful!

Dataset Information

• Dataset DOI: https://doi.org/10.5683/SP3/QF1VTK
• Dataset Created: 2023-09-17
• Created by: Kehan (Sky) Sheng

Contributors

• Principal Investigator: Daniel Weary

  • ORCID: 0000-0002-0917-3982
  • Affiliation: University of British Columbia
  • Email: [email protected]

• Co-investigator: Marina von Keyserlingk

  • ORCID: 0000-0002-1427-3152
  • Affiliation: University of British Columbia
  • Email: [email protected]

• Co-investigator: Tiffany-Anne Timbers

  • ORCID: 0000-0002-2667-376X
  • Affiliation: University of British Columbia
  • Email: [email protected]

• Contributor: Kehan (Sky) Sheng

  • ORCID: 0000-0001-6442-5284
  • Affiliation: University of British Columbia
  • Email: [email protected]

• Contributor: Borbala Foris

  • ORCID: 0000-0002-0901-3057
  • Affiliation 1: University of British Columbia
  • Affiliation 2: University of Veterinary Medicine, Vienna
  • Email: [email protected]

• Contributor: Varinia Cabrera

  • ORCID: 0009-0007-7819-6612
  • Affiliation 1: University of British Columbia
  • Affiliation 2: University of the Republic, Uruguay
  • Email: [email protected]

Project Information

• Date of Video Collection: March 30, 2021 - June 11, 2021
• Location of Video Collection: UBC Dairy Education and Research Centre, 6947 No. 7 Highway, Agassiz, BC V0M 1A0, Canada
• Funding: This project was supported by the NSERC Industrial Research Chair, University of British Columbia Land and Food System Internal Research Grant.