Dog Breed Classifier

Goal

• Create a Neural Network model that able to predict Dog breed based on sample images
• Develop a simple web application where user can upload an image and receive predictions

Dataset

I use Stanford Dogs Dataset as the main source of samples. It contains around 12 thousands of dog images into 120 breeds.

Model

I use only MobileNetV2 with some fine-tuning. The result is very small (around 39MB) yet pretty accurate model. You can find out more about the model training in the notebooks directory.

Performance

The model achieve overal accuracy of 78%. Maybe the result is pretty biased because of the lack of test samples.