A database of dentition images of Indian breed cattle and estimation of cattle’s age using deep learning algorithms

Accurate age estimation in livestock is crucial in agricultural management, influencing breeding, healthcare, and production planning. The primary method entails visually inspecting the teeth, a procedure susceptible to human fallibility. Recent deep learning (DL) algorithms are highly efficient in...

Full description

Saved in:
Bibliographic Details
Published in:Engineering applications of artificial intelligence Vol. 162; p. 112172
Main Authors: Patil, Chinmay Vijay, Bhurane, Ankit Ashokrao, Ghasad, Preeti, Kamble, Vipin, Sharma, Manish, Singh, Anand, Nandeshwar, Nareshkumar, Tan, Ru-San, Acharya, Rajendra
Format: Journal Article
Language:English
Published: Elsevier Ltd 24.12.2025
Subjects:
Automated age estimation
Convolutional neural networks
Deep learning
Instance segmentation
Livestock management
Vision transformer
Deep learning
Instance segmentation
Vision transformer
Convolutional neural networks
Automated age estimation
Livestock management
ISSN:0952-1976
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Accurate age estimation in livestock is crucial in agricultural management, influencing breeding, healthcare, and production planning. The primary method entails visually inspecting the teeth, a procedure susceptible to human fallibility. Recent deep learning (DL) algorithms are highly efficient in automating age prediction based on dentition images, offering faster and more accurate results. To our knowledge, no dataset exists for Indian livestock, hindering DL algorithm development for teeth segmentation and age estimation. Therefore, we created a database of dentition images for Indian livestock, covering an age range of 1 to 15 years, using 2,883 subjects. In this research, we employed a hybrid model based on You Only Look Once version 8 (YOLOv8) for instance segmentation of raw dentition images. Segmentation achieved 73.59% mean Average Precision (mAP) 50 and 56.44% mAP50–90, indicating strong performance in varying localization thresholds. These segmented images were then used in various DL algorithms, including convolutional neural network (CNN)-based transformers and vision transformers (ViTs), to develop DL models for age estimation. We conducted three age estimation experiments with increasing granularity: four age groups (4-year intervals), seven age groups (2-year intervals), and fifteen age groups (1-year intervals). InceptionV3 achieved the highest accuracy in the first two experiments with 73.72% and 62.41%, respectively. In the third experiment, MobileNet performed best with an accuracy of 43.87%. Our results demonstrate the effectiveness of these intricate DL models in enhancing livestock age estimation, making the process more streamlined and reliable. The study may assist farmers in making informed and efficient agricultural management decisions. •Dataset: 2,883 Indian cattle dental images (ages 1–15) curated for age estimation.•Method: AI framework using segmentation, CNNs and transformers estimates cattle age.•Impact: Non-invasive AI dentition analysis enhances livestock age assessment.•Visualization: Saliency maps reveal key dental features that guide the AI model’s decisions.
ISSN:0952-1976
DOI:10.1016/j.engappai.2025.112172