Used the Oxford-IIIT Pet Dataset and some pictures of my cats to fine-tune a Resnet34 model to classify images into three categories:

  1. Bytes (my older white and grey cat)
  2. Ganami (my younger tabby cat)
  3. Some other cat

I did this as a learning experiment while taking the amazing Practical Deep Learning for Coders course. It is truly one of the best courses I have ever taken. I absolutely love this course.

I Used about 2000 images of random cats, 35 images of my cat Bytes, and 35 images of my other cat Ganami.

Surprisingly I was able to get pretty good accuracy with a single epoch of training:

Screenshot 2025-07-09 at 7.27.08 PM.png

I’m an Android dev and I’m super excited about Google AI Edge so naturally I decided to take this model and convert it to Tensorflow Lite and load it onto Android. I built a simple app which allows to take pictures and show the model’s predictions.

Here are some screenshots of the app in action:

Screenshot_20250709-123113.png

Screenshot_20250709-123308.png

Screenshot_20250709-123549.png