A project to build everything around garbage classification advisor.
构建一个生活垃圾分类助手。
.
├── gc_api # Backend RESTful APIs, based on Flask.
├── simple-citizen # Frontend interface, based on React.
├── category # Storing categories of cities and graph related services.
├── chatboxes # Garbage classification chatbot.
├── models # Models and models configuration file folder.
├── notebooks # Jupyter notebooks for oneoff actions (images pre-processing, model training, etc).
└── spiders # Data collector for collecting images of garbages.
Browser may prompt warning saying this site is not secure, you can click "Advanced" and proceed to visit the site.
https://122.51.49.45
- Currently, the taxonomy is base on Shanghai guideline. 4 kinds of domestic rubbish, which are [可回收物, 有害垃圾, 湿垃圾, 干垃圾].
- A taxonomy on Guangzhou guideline is coming.
- Model is building on transfer learning based on Inception V3 and MobileNet V2.
- Current Accuracy is around 80%.
- Currently it's using top layers as classifier labels. Next is to train a model on recognize the object itself, such as wasted paper, wasted can, food waste, etc.
- A Flask based RESTful backend, accepting image and return a classification result.
- A React based frontend, serving the API as a Web application. Allowing user to take a photo and upload and display the result.