Dog Breed Classifier with Heroku deployment

This is a project which detects 30 different dog breeds.

Key features are :

This model is trained on Tensorflow 2.3.0 object classification.
To train the model, famous Kaggle Dataset which consists of 120 different breeds, out of which 30 famous breeds were taken for simplicity.
Took nearly 2.5k images of 30 breeds.
For training,Inception v3 model as base model is used for feature extraction.
After training, training accuracy reached 99% with testing accuracy nearing 94%.
Deployed the model with python and having 'Flask' as backend and 'gunicorn' as server.

Visit dog-breed-classification website with This link .

Detailed training is explained in final_building_model.ipynb file.

Trained model 'smart_model_mobile.h5' is available in google drive with this link

Snapshot of the website's Front page and Breed page where name of breeds which model can predict are mentioned.

This is the prediction page before uploading and after uploading and getting predictions:-

Feel free to contact me incase on any quaries

Yashodhara Shastri G
E-mail : [email protected]