Skip to content

A fun program that uses ResNet-50 with bottleneck features and Haar-Cascade Face Detection to determine dog breed.

Notifications You must be signed in to change notification settings

treble-maker123/dog-breed-classifier

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

11 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

This repository contains code for training a dog breed classifier. A demo of this app can be found here.

This model is built on top of the ResNet-50 developed by Kaiming He et al. A GlobalAveragePooling2D layer and a fully-connected layer with softmax activation is added as trainable layers. The model is then trained with 6680 additional images of 133 breeds of dogs.

The original training (current demo) did not incorporate data augmentation, and the accuracy from validating on the test set was around 80%. The train.py script does have data augmentation code, and I will train a new model with it in the near future and update the demo with the new model.

If you would like to try this out yourself, feel free to reach out to me for the dog images.

About

A fun program that uses ResNet-50 with bottleneck features and Haar-Cascade Face Detection to determine dog breed.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages