项目作者： cyyeh
项目描述 ：
dog breed classification project
高级语言： JavaScript
项目地址: git://github.com/cyyeh/dog-breed-classifier-project.git
Dog Breed Classification Project
Project Functionalities
- Users can know more about dog breeds and their information through taking photos or uploading photos.
- Users can also directly use photos provided by the project to quickly play around.
- Users can get top 3 dog breed prediction results with wikipedia links.
- Users can learn English and Chinese at the same time.
System Architecture
- dog detection
- model: MobileNet V2 pretrained model without finetuning
- input: an image having RGB channels with Base64 encoding
- output: boolean
- dog classification
- model: MobileNet V2 pretrained model with finetuning
- input: an image having RGB channels with Base64 encoding
- output: top-3 prediction results(dog breed names and prediction probabilities)
Project Structure
- api: backend
- web: frontend
- frontend
- HTML/CSS/JavaScript
- Materialize
- Babel
- Webpack
- backend
- deep learning
Deep Learning Under the Hood in This Project
You can check this repo for further details, and this project is orginally from my Udacity Machine Learning Engineer Nanodegree capstone project. However, I have changed some model details for this dog breed classifier, you can check here for more details.
Setup
Frontend
please see the web folder
npm install (download dependencies first)- local development
- deployment
You can checkout other commands in package.json in the web folder
Backend
please see the api folder
- local development without docker:
pip install pipenv (If you don’t have pipenv installed)pipenv installpipenv run uvicorn main:app --reload
Below commands are already written in Makefile, please checkout and change any setting you need first.
- local development using docker:
- build a docker image:
- submit a docker image to Google Container Registry:
- deploy a docker image in Google Container Registry to Google Cloud Run:
Todos
