bangkit-machine-learning

The source code of machine learning model's API of Travens smart guide in order to complete Bangkit Capstone Project

Model Building's Notebook

Landmark Detection Notebook

API URL

Flask API

Dataset Resources

• Google Landmarks Dataset V2
• Traveloka
• Google Images
• Wikipedia

Detection Labels

Label Map

API Endpoint

Endpoint	Method	Body Sent (JSON)	Description
/	GET	None	HTTP GET REQUEST Testing Endpoint
/post	POST	Anything	HTTP POST REQUEST Testing Endpoint
/404	GET & POST	Anything	404 NOT FOUND Endpoint
/predict	POST	file: Image file	HTTP POST REQUEST Prediction Endpoint

The flow of Machine Learning Service

How to run this Flask app

• Clone this repo
• Open terminal and go to this project's root directory
• Type python -m venv .venv and hit enter
• In Linux, type source .venv/bin/activate
• In Windows, type .venv\Scripts\activate
• Type pip install -r requirements.txt
• Serve the Flask app by typing flask run
• It will run on http://127.0.0.1:5000

How to predict image with Postman

• Open Postman
• Enter URL request bar with http://127.0.0.1:5000/predict
• Select method POST
• Go to Body tab and select form-data
• Change key from form-data with file (it must be named file)
• Input the image that you want predict as a value of the key
• Send the request

Architecture of SSD MobileNet V2 for Landmark Detection

References

• Cao B, Araujo A, Sim J. Unifying Deep Local and Global Features for Image Search. 2020. Arxiv
• Weyand T, Araujo A, Cao B, Sim J. Google Landmarks Dataset v2-A Large-Scale Benchmark for Instance-Level Recognition and Retrieval. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020. Arxiv
• A Howard, A Zhmoginov, LC Chen, M Sandler, M Zhu. MobileNetV2: Inverted Residuals and Linear Bottlenecks. The IEEE Conference on Computer Vision and Pattern Recognition. 2018. Arxiv
• Collection of models for landmark recognition TensorFlow Hub