PawtectAI

AI-driven image analysis for quick detection of common pet health conditions.

Comment

Inspiration

Many pet owners struggle to identify early signs of health issues like skin infections, eye diseases, or ear infections. Without timely detection, these conditions can worsen and cause unnecessary suffering. We wanted to use AI to make pet care more accessible and help owners spot potential health problems earlier.

What it does

PawtectAI is an AI-powered web app that detects common health issues in pets from images. Users upload a photo of their pet. The AI model analyzes it and classifies it into one of four categories: ✅ Healthy ⚠️ Skin Infection ⚠️ Eye Disease ⚠️ Ear Infection This helps pet owners get a quick AI-powered health check and decide if they need to consult a veterinarian.

How we built it

Collected and organized a dataset of pet images across four classes (Healthy, Skin Infection, Eye Disease, Ear Infection). Used transfer learning with MobileNetV2 to achieve higher accuracy with limited data. Applied data augmentation to improve generalization. Built a simple Streamlit interface where users can upload images and see predictions instantly. Challenges we ran into Finding enough high-quality images for each health condition. Avoiding overfitting with a relatively small dataset. Designing an interface simple enough for all pet owners to use. Accomplishments we’re proud of Built a functional AI health detection tool in a short timeframe. Created something meaningful that could assist pet owners in caring for their animals. Learned how to combine AI + web apps to solve a real-world problem.

What we learned

How to apply transfer learning for multi-class classification. How to deploy an AI model into a user-friendly app. That AI can empower pet owners with better health insights.

What’s next for PawtectAI

Expand the dataset with more diverse images for better accuracy. Add support for detecting additional pet health conditions. Build a mobile app for easier access. Collaborate with veterinarians to validate and refine the tool.

Built With

  • flask-(or-streamlit)
  • google-colab
  • mobilenetv2
  • python
  • tensorflow/keras
Share this project:

Updates

posted an update —

1.Just trained the first version of our MobileNetV2 model on 4 classes (Healthy, Skin Infection, Eye Disease, Ear Infection). Early results show ~60% accuracy on the validation set. Excited to optimize further! 2.PawtectAI is now live as a demo! You can try it out directly in Colab here: [Colab Demo Link] 3.Fine-tuned the model by unfreezing MobileNetV2 layers and training with a lower learning rate. Accuracy improved to 92%!

Log in or sign up for Devpost to join the conversation.