UCHEGBU, PROMISE EKENE (2026) DEVELOPMENT OF AN AI-POWERED WEB-BASED CATTLE LAMENESS DETECTION SYSTEM BY. Other thesis, Godfrey Okoye University, Enugu Nigeria.
![[img]](https://eprints.gouni.edu.ng/style/images/fileicons/text.png) |
Text
PDF .pdf Download (1MB) |
Abstract
Lameness in cattle is one of the main problems for both dairy and beef farming since it causes decreased milk yield, reduced fertility rate, increased costs for veterinary services, and premature slaughtering. Detecting the problem is hard even on some farms due to the subjective nature of visual assessments used to detect lameness. Therefore, this paper presents the results of the development of an internet-based automated cattle lameness detection system based on video analysis using computer vision techniques and machine learning algorithms. To detect and track cattle in video images, this system employs a pre- trained YOLOv8 model. The system identifies several gait parameters such as movement velocity, stride displacement, straightness of the path, lateral swaying, bounding box variation ratio, vertical oscillations, stride regularity, and lateral imbalance, which serve as input values for the machine learning model. An SVM model trained using 50 freely accessible cattle videos split into 80% training and 20% testing samples scored an accuracy of 70%. The output from the classifier showed that the classes have asymmetric performance, where the lame class has a precision of 0.75, a recall of 0.60, and an F1 score of 0.67 while the normal class has a precision of 0.67, a recall of 0.80, and an F1 score of 0.73. The F1 score generated for the classes resulted in a macro-average F1 score of 0.70. Therefore, the system was capable of recognizing gait differences but still needs more improvement to be able to consistently predict lame instances. The application was implemented using Flask technology, allowing users to automatically input cattle videos into the system for lameness diagnosis along with the probability of lameness in form of severity. This study showed that it is feasible to use gait analysis through videos as a cheap decision-support tool in diagnosing lameness cases. Since the model was trained using a limited dataset, future research should aim at improving the system by training it using a diverse dataset in terms of breeds, cameras used, environment, and lameness severity.
| Item Type: | Thesis (Other) |
|---|---|
| Subjects: | Q Science > Q Science (General) |
| Divisions: | Faculty of Computing And Information Technology (FACIT) |
| Depositing User: | Cynthia Ugwuoti |
| Date Deposited: | 24 Jul 2026 13:54 |
| Last Modified: | 24 Jul 2026 13:54 |
| URI: | http://eprints.gouni.edu.ng/id/eprint/5993 |
Actions (login required)
 |
View Item |
