Page 105 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

Use Case 12: AI-Based Horse Identification and Gait Analysis�

Organization: AIforPet 4.1-Healthcare

Country: South Korea

Contact Person(s):

Primary contact: Euna Huh, eahuh@ aiforpet .com
Secondary contact: Seonghoon Kim, shkim@ aiforpet .com
Any other contacts: Jihyun Seo, jhseo@ aiforpet .com

1 Use Case Summary Table

Item Details

Category Healthcare

Problem Addressed Horses, known for their close bond with humans, are frequently exposed
to health risks such as lameness, which in severe cases can lead to the loss
of the animal. Horse owners strive to minimize these risks and seek ways
to detect problems as early as possible. At the same time, veterinarians
are in need of additional tools that can enhance diagnostic objectivity
and support more systematic and proactive health management.
Key Aspects of Solu- Image segmentation, Pose Estimation, Gait Analysis, Image Processing,
tion Mobile apps

Technology Horse Identification, Horse Lameness Detection, Animal Healthcare
Keywords

Data Availability Private; Internal data used for model training and validation
Metadata (Type of Image data, Video data, Annotation (Segmentation, Pose Estimation) data
Data)
Model Training and • Real-time Multi-Person Pose Estimation (RTMPose) for pose estimation
Fine-Tuning • Segmentation Transformer (SegFormer) and Mask Region-Based
Convolutional Neural Network (Mask R-CNN) for horse passport
generation
• Motion Contrastive Anomaly Detection (MoCoDAD) and Convolu-
tional Neural Network–Long Short-Term Memory (CNN-LSTM) models
for lameness detection

Testbeds or Pilot The system has been evaluated in real-world settings through pilot
Deployments deployments, demonstrating its practical feasibility and user acceptance
[1].

Code repositories Private

69