Overview

This repository provides search experience with your images by Azure Computer Vision. Using streamlit framework, we can easily build an app, and customize it with your preferences.

Prerequisites

• Python environment
• Azure subscription
  • Computer vision resource
• Your image files with jpg, jpeg or png format¹

How to build an app

1. Install necessary libraries

• Please execute the following in your terminal:

  pip install -r ./requirements.txt

2. Configure the environment

• By using common/config.ini, you can specify the setting. See this page in detail.

  [Azure]
  subscription_key = 12345xxxyyy
  cv_endpoint = https://xxxx.cognitiveservices.azure.com/
  language = ja

3. Run streamlit app with your terminal:

• Run the app

  streamlit run ./st_app.py

• You will find the following message. Then, copy and pate the URL in your browser²:

  You can now view your Streamlit app in your browser.
  Local URL: http://localhost:8501
  Network URL: http://xxx.xxx.xxx.xxx:8501
  

• Once you successfully run your app, you will see:

How to use the app

1. Register image

• In registering your image, select your image in your local folder

• Once you select your image, you will find the analyzed result³:

2. Search image

• If you input your interest by word, your will find the search results:

3. Customize your preferences

If you want to more results, change the prarmeters:

• topN: 5 tags are shown in each picture

• Search result: 4 search results are shown

4. Further usage

• If you want to generate search index with your images, follow the instructions:
  • Delete df_tag_db.pkl file, and all images from upload_images directories.
  • Follow the instruction 1. Register image.
• IF you ignore SSL check in consuming Azure CV, set SSL_check variable as False in st_app.py. It's not recommended, but please change it under your environment.

Directory structure

├─common
│  ├─config.ini         : Configuration file. Generate it by yourself
│  └─generate_tag_db.py : Main functions
├─docs                  : Defines this repository
├─upload_images         : Store uploaded images
│  └─images             : Your images
├─ df_tag_db.pkl        : Defines search index as pandas DataFrame
├─ st_app.py            : App in streamlit
└─ README.md            : Instruction of the repository

References

• Azure Computer Vision
  • Quick Start: Image analysis with Computer Vision
  • Call the Image Analysis API in Azure Computer Vision
  • computer_vision_samples in github
• Streamlit
  • How to build stream app

Footnotes

1. Refer to detailed requirements for images ↩

2. Go to streamlit doc for more reference. ↩

3. As we see, the presented tags are in Japanese. If you want to change the displayed language, change language in common/config.ini as en, for example. Go here, if you need more detail. ↩