ml-image-searcher

NodeJS module NPM package that helps to search for images using instagram hashtags and machine learning object recognition technology.

Disclaimer 1

This package was created in 2019 and was depending on ?__a=1, at that time instagram allows the usage of /?__a=1&is_video=true but now instagram no longer allows this and also not allows fetch profiles info without auth session.
Maybe in the future I will implement a scrapping-based solution for this package.

Disclaimer 2

Instagram has gone to great lengths to prevent scraping and other unauthorized access to their public content. This module is dependant on the markup the public-facing instagram.com. Should that change this module might also stop working as intended. You should take this into consideration when deciding whether this module will work for you.

Motivation

In example below output shows how searching using:

• Human hashtags + ML: 20(cats) / 20(total result) -> [100% correct]
  
• Human hashtags : 12(cats) / 20(total result) -> [60% correct]

Installation

npm i ml-image-searcher

Usage

note: there are some limitation of loading instagram data, but it still usefull in most cases :)

Search for cats photos using human hashtags + machine learning prediction

const IG = require('ml-image-searcher');

IG.getImages('cat').then((ret) => console.log(ret.filteredData))

Output (format: htmlAll) response:

• Left: Human hashtags + ML
• Right: Human hashtags only

Output (format: filteredData) response:

{ "node":
     { "comments_disabled": false,
       "__typename": "GraphImage",
       "id": "2166483334784860896",
       "edge_media_to_caption": "[Object]",
       "shortcode": "B4Q5R1Np8Lg",
       "edge_media_to_comment": "[Object]",
       "taken_at_timestamp": 1572484960,
       "dimensions": "[Object]",
       "display_url":
        "https://instagram.fatz1-1.fna.fbcdn.net/vp/a7d0f86a15a5d4830f08263741c057ec/5E3E0F4B/t51.2885-15/e35/76909697_2616077041746584_1323244633407869990_n.jpg?_nc_ht=instagram.fatz1-1.fna.fbcdn.net&_nc_cat=108",
       "edge_liked_by": "[Object]",
       "edge_media_preview_like": "[Object]",
       "owner": "[Object]",
       "thumbnail_src":
        "https://instagram.fatz1-1.fna.fbcdn.net/vp/620528b4197d036abda248ccbe36697e/5E61A8F1/t51.2885-15/sh0.08/e35/s640x640/76909697_2616077041746584_1323244633407869990_n.jpg?_nc_ht=instagram.fatz1-1.fna.fbcdn.net&_nc_cat=108",
       "thumbnail_resources": "[Array]",
       "is_video": false,
       "accessibility_caption": "Image may contain: cat" } },
  { "node":
     { "comments_disabled": false,
       "__typename": "GraphImage",
       "id": "2166483275533936833",
       "edge_media_to_caption": "[Object]",
       "shortcode": "B4Q5Q-CBmjB",
       "edge_media_to_comment": "[Object]",
       "taken_at_timestamp": 1572484953,
       "dimensions": "[Object]",
       "display_url":
        "https://instagram.fatz1-1.fna.fbcdn.net/vp/bd226ff8cf673943e32aa28696a9e66e/5E64A3DA/t51.2885-15/e35/73284628_148613809747256_665998994713792293_n.jpg?_nc_ht=instagram.fatz1-1.fna.fbcdn.net&_nc_cat=108",
       "edge_liked_by": "[Object]",
       "edge_media_preview_like": "[Object]",
       "owner": "[Object]",
       "thumbnail_src":
        "https://instagram.fatz1-1.fna.fbcdn.net/vp/3ff84475c7f81dc9c1e75d6524609b65/5E48B2A9/t51.2885-15/sh0.08/e35/s640x640/73284628_148613809747256_665998994713792293_n.jpg?_nc_ht=instagram.fatz1-1.fna.fbcdn.net&_nc_cat=108",
       "thumbnail_resources": "[Array]",
       "is_video": false,
       "accessibility_caption": "Image may contain: cat" } },
  ...
}

---

Run Locally:

1. clone this repo
2. cd ml-image-searcher/
3. change LISTENER variable in .env file to: LISTENER='true'
4. run node index.js
5. open browser: localhost:8080

Response formats:

{
    originalData,      // result based on humans hashtags only
    filteredData,      // result based on humans hashtags and ML prediction
    htmlOriginalData,  // originalData images only in html format
    htmlFilteredData,  // filteredData images only in html format
    htmlAll,           // both results (htmlOriginalData + htmlFilteredData) in a table
};