This is a AutoML based Image Classification model that is hosted on Flask, allowing users to upload doggy photos and see which category of doggy their photos belong to.
This web application is deployed with Kubernetes and has passed the load test of 1k+ requests served by multiple endpoints per second.
https://www.youtube.com/watch?v=1fUn561it5k
It is possible that a "FailedPrecondition: 400" will occur when requesting a classification result. That means the Google AutoML model deployment has been turned off (The billing of this Google AutoML is time-based and it's quite costly).
- Create a virtualenv and source
- Run
make install
- Run in local:
python web.py - Run in Docker:
./run_docker.sh
Our dataset comes from Standford Dog Breed Identification on Kaggle, which has 120 dog breeds and ~150 images per class. Considering the long running time and expensive cost of this whole dataset if trained with AutoML, we select 3 classes of dogs, Dingo, Boxer, Samoye and 20 images per class from it.
The Image Classification model is build with AutoML.
Here are the model and its performance displays.
The AutoML based Image Classification model is hosted on Flask, allowing users to upload doggy photos and see which category of doggy their photos belong to.
Docker Image is built and upload to Docker Hub for deployment preparation.
This web application is deployed to Google Kubernetes Engine and has passed the load test of 1k+ requests served by multiple endpoints per second.
| Name | Github Username |
|---|---|
| Zhihao Shu | harryshu1997 |
| Bodong Wang | bow017 |
| Rui Xu | Ruixxxx |