Skip to content

Latest commit

 

History

History
85 lines (65 loc) · 1.79 KB

README.md

File metadata and controls

85 lines (65 loc) · 1.79 KB

image-tagger

An image tagger using Groq. Use GroqChat to help tag your images into a json file.

Powered by Groq for fast inference.

Prerequisites

Go to Groq. Go to the dev console and make an API Key. Create a .env file in the root of the repository like so, replacing the XXXXX with your value.

GROQ_API_KEY=XXXXX

Run the following from the root of the repository to get a functional conda environment.

conda env create -f environment.yaml

Usage

Please note that this is not perfect. Groq sometimes returns invalid data.

python main.py -i "./tests/images/800px-Another_Day_in_NYC_(2539547867).jpg" "https://upload.wikimedia.org/wikipedia/commons/thumb/1/17/Interior_Galeria_Arte_Nacional_%28Caracas%29.jpg/800px-Interior_Galeria_Arte_Nacional_%28Caracas%29.jpg?20071010223618"

Example (output.json)

{
  "./tests/images/800px-Another_Day_in_NYC_(2539547867).jpg": {
    "tags": [
      "Flowers",
      "Nature",
      "Pink",
      "Potted",
      "Green",
      "Plant"
    ],
    "categories": [
      "Gardening",
      "Flowers"
    ]
  },
  "https://upload.wikimedia.org/wikipedia/commons/thumb/1/17/Interior_Galeria_Arte_Nacional_%28Caracas%29.jpg/800px-Interior_Galeria_Arte_Nacional_%28Caracas%29.jpg?20071010223618": {
    "tags": [
      "fountain",
      "art",
      "sculpture",
      "relax",
      "vietnam"
    ],
    "categories": [
      "Museum",
      "Culture"
    ]
  }
}
python main.py -i ./tests/images

Images

All images were source from Wikimedia commons.

Tests

Run the following from the root

tox