spctr01 / Doogle

A web-based dog breed identifier in PyTorch deployed using Flask. This is an example of transfer learning for which Resnet152 is used.

Deploying model to web.

Front page

Shows predicted breed and information about breed

About

This is example of transfer learning. Deployed using Flask frontend is designed using css, javascript . vgg19(pretrained on imagenet) to find wether image contains dog or not and resnet152 to find the breed of dog. dataset of 133 dog breeds is used to train the model

Dog breed Dataset

• models folder contains resent and vgg model weight files instead of read.txt

Installation & usage

Download or clone the repository by :

git clone https://github.com/spctr01/dog_breed.git

move into folder:

 cd dog_breed

Install the requirements:

pip install -r requirements.txt

Download the model weight files and paste the files (vgg.pth & model.pt) to Models Folder

Resnet152

Vgg19 (change name to vgg.pth)

run the commands(running flask app):

export FLASK_APP=app.py
flask run