Relic / Monument / Landmark Classifier for Dinara

Bangkit 2022 Capstone Project Classifier for Dinara.

The models can currently classify only 29 classes but can be expanded in the future. More info about relic / monument / landmark that can be classified can be found in the notebooks.

Application Architecture

More Info on Model Creation

Note that we haven't published our dataset yet, but we are working on it

Xception Architecture

Notebook

Usage

Usage is really straight up, you can either use our deployed model right away or deploy the model in your own Google Cloud Platform

Deploying Yourself

1. Create new Google Cloud project
2. Install Cloud Run API && Cloud Build API
3. Install and init Google Cloud SDK
4. Run commands below

gcloud builds submit --tag gcr.io/relic-classifier/index
gcloud run deploy --image gcr.io/relic-classifier/index --platform managed

after deploying model with steps above we simply send a file using HTTP POST with 'imagefile' and the images itself as key value pair and the deployed model sends back a JSON response containing the predicted image.

import requests

resp = requests.post("https://getprediction-qyqf4nfema-et.a.run.app", files={'imagefile': open('./test/example_images/candi-borobudur.jpg', 'rb')})

print(resp.json())

or you can just use it locally by running 'main.py' and flask will run a local server instead

import requests

resp = requests.post("http://127.0.0.1:5000/", files={'imagefile': open('./test/example_images/candi-borobudur.jpg', 'rb')})

print(resp.json())