CattleGuard AI

• 

  Default Screen

• 

  Uploaded Image

• 

  Disease Prediction Graph

• 

  Input Image

Inspiration

Cattle skin diseases are often detected late due to limited veterinary access. We wanted to enable instant, photo-based diagnosis to reduce losses and improve animal health.

What it does

CattleGuard AI uses a MobileNetV2-based model to detect cattle skin diseases and severity from images, providing instant predictions to farmers and vets.

How we built it

We used transfer learning with MobileNetV2, applied image preprocessing and augmentation, handled class imbalance, and trained the model using TensorFlow/Keras for real-time inference.

Challenges we ran into

Inconsistent image quality, class imbalance, and avoiding overfitting while ensuring real-world generalization.

Accomplishments that we're proud of

Building a lightweight, deployable end-to-end AI system suitable for on-field use.

What we learned

The importance of data quality, transfer learning, and designing ML systems for deployment, not just accuracy.

What's next for CattleGuard AI

Expanding the dataset, adding explainability (Grad-CAM), and deploying an offline-capable mobile app.

Built With

• git
• github
• google-colab
• jupyter-notebook
• keras
• matplotlib
• mobilenetv2
• numpy
• opencv
• pandas
• python
• scikit-learn
• streamlit
• tensorflow