DEV Community

Cover image for Scrape and download Google Images with Python
Dimitry Zub
Dimitry Zub

Posted on • Updated on • Originally published at serpapi.com

Scrape and download Google Images with Python

Contents: intro, imports, what will be scraped, process, code, links, outro.

Intro

This blog post is a continuation of Google's web scraping series. Here you'll see how to scrape Google Images using Python with beautifulsoup, requests, lxml libraries. An alternative API solution will be shown.

Prerequisites: basic knowledge of beautifulsoup, requests, lxml and regular expressions.

Imports

import requests, lxml, re, json, urllib.request
from bs4 import BeautifulSoup
Enter fullscreen mode Exit fullscreen mode

What will be scraped

Suggested search results
image

Image results
image

Process

Firstly, selecting container with needed data, name, link, chips for suggested search results. CSS selector reference.

Secondly, looking where decoded thumbnail URLs are located for suggested search results.

On the GIF above I copied data:image part from src attribute to check with Chrome find bar (Ctrl+F) if something I was looking for is located in the <script> tags.

Why can't I just parse src attribute from img element?

If you parse <img> with src attribute, you'll get an 1x1 placeholder instead of actual thumbnail.

The logic was:

  1. find where in the <script> tags decoded thumbnail URLs were located.
  2. split certain chuck of the text from the big blob of text to a smaller one for easier regex extraction.
  3. create a regex pattern that will match all URLs. The same process was applied to extract original resolution images with their thumbnails.

To test your regular expression you can use regex101.

Save images
To save images you can use urllib.request.urlretrieve(url, filename) (more in-depth)

# often times it will throw 404 error, so to avoid it we need to pass user-agent

opener=urllib.request.build_opener()
opener.addheaders=[('User-Agent','Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/70.0.3538.102 Safari/537.36 Edge/18.19582')]
urllib.request.install_opener(opener)

urllib.request.urlretrieve(original_size_img, f'LOCAL_FOLDER_NAME/YOUR_IMAGE_NAME.jpg')
Enter fullscreen mode Exit fullscreen mode

Code

Start of the code for both functions

import requests, lxml, re, datetime
from bs4 import BeautifulSoup

headers = {
    "User-Agent":
    "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/70.0.3538.102 Safari/537.36 Edge/18.19582"
}

params = {
    "q": "minecraft shaders 8k photo",
    "tbm": "isch",
    "ijn": "0",
}

html = requests.get("https://www.google.com/search", params=params, headers=headers)
soup = BeautifulSoup(html.text, 'lxml')
Enter fullscreen mode Exit fullscreen mode

Scrape Suggested search results: name, link, chips, thumbnail.

def get_suggested_search_data():
    for suggested_search in soup.select('.PKhmud.sc-it.tzVsfd'):
        suggested_search_name = suggested_search.select_one('.hIOe2').text
        suggested_search_link = f"https://www.google.com{suggested_search.a['href']}"

        # https://regex101.com/r/y51ZoC/1
        suggested_search_chips = ''.join(re.findall(r'=isch&chips=(.*?)&hl=en-US', suggested_search_link))
        print(f"{suggested_search_name}\n{suggested_search_link}\n{suggested_search_chips}\n")

    # this steps could be refactored to a more compact
    all_script_tags = soup.select('script')

    # https://regex101.com/r/48UZhY/6
    matched_images_data = ''.join(re.findall(r"AF_initDataCallback\(({key: 'ds:1'.*?)\);</script>", str(all_script_tags)))

    # https://kodlogs.com/34776/json-decoder-jsondecodeerror-expecting-property-name-enclosed-in-double-quotes
    # if you try to json.loads() without json.dumps() it will throw an error:
    # "Expecting property name enclosed in double quotes"
    matched_images_data_fix = json.dumps(matched_images_data)
    matched_images_data_json = json.loads(matched_images_data_fix)

    # search for only suggested search thumbnails related
    # https://regex101.com/r/ITluak/2
    suggested_search_thumbnails_data = ','.join(re.findall(r'{key(.*?)\[null,\"Size\"', matched_images_data_json))

    # https://regex101.com/r/MyNLUk/1
    suggested_search_thumbnail_links_not_fixed = re.findall(r'\"(https:\/\/encrypted.*?)\"', suggested_search_thumbnails_data)

    print('Suggested Search Thumbnails:')  # in order
    for suggested_search_fixed_thumbnail in suggested_search_thumbnail_links_not_fixed:
        # https://stackoverflow.com/a/4004439/15164646 comment by Frédéric Hamidi
        suggested_search_thumbnail = bytes(suggested_search_fixed_thumbnail, 'ascii').decode('unicode-escape')
        print(suggested_search_thumbnail)


get_suggested_search_data()
Enter fullscreen mode Exit fullscreen mode

Scrape Google images: title, link, source, thumbnail, original resolution image (and download them):

def get_images_data():

    print('\nGoogle Images Metadata:')
    for google_image in soup.select('.isv-r.PNCib.MSM1fd.BUooTd'):
        title = google_image.select_one('.VFACy.kGQAp.sMi44c.lNHeqe.WGvvNb')['title']
        source = google_image.select_one('.fxgdke').text
        link = google_image.select_one('.VFACy.kGQAp.sMi44c.lNHeqe.WGvvNb')['href']
        print(f'{title}\n{source}\n{link}\n')

    # this steps could be refactored to a more compact
    all_script_tags = soup.select('script')

    # # https://regex101.com/r/48UZhY/4
    matched_images_data = ''.join(re.findall(r"AF_initDataCallback\(([^<]+)\);", str(all_script_tags)))

    # https://kodlogs.com/34776/json-decoder-jsondecodeerror-expecting-property-name-enclosed-in-double-quotes
    # if you try to json.loads() without json.dumps it will throw an error:
    # "Expecting property name enclosed in double quotes"
    matched_images_data_fix = json.dumps(matched_images_data)
    matched_images_data_json = json.loads(matched_images_data_fix)

    # https://regex101.com/r/pdZOnW/3
    matched_google_image_data = re.findall(r'\[\"GRID_STATE0\",null,\[\[1,\[0,\".*?\",(.*),\"All\",', matched_images_data_json)

    # https://regex101.com/r/NnRg27/1
    matched_google_images_thumbnails = ', '.join(
        re.findall(r'\[\"(https\:\/\/encrypted-tbn0\.gstatic\.com\/images\?.*?)\",\d+,\d+\]',
                   str(matched_google_image_data))).split(', ')

    print('Google Image Thumbnails:')  # in order
    for fixed_google_image_thumbnail in matched_google_images_thumbnails:
        # https://stackoverflow.com/a/4004439/15164646 comment by Frédéric Hamidi
        google_image_thumbnail_not_fixed = bytes(fixed_google_image_thumbnail, 'ascii').decode('unicode-escape')

        # after first decoding, Unicode characters are still present. After the second iteration, they were decoded.
        google_image_thumbnail = bytes(google_image_thumbnail_not_fixed, 'ascii').decode('unicode-escape')
        print(google_image_thumbnail)

    # removing previously matched thumbnails for easier full resolution image matches.
    removed_matched_google_images_thumbnails = re.sub(
        r'\[\"(https\:\/\/encrypted-tbn0\.gstatic\.com\/images\?.*?)\",\d+,\d+\]', '', str(matched_google_image_data))

    # https://regex101.com/r/fXjfb1/4
    # https://stackoverflow.com/a/19821774/15164646
    matched_google_full_resolution_images = re.findall(r"(?:'|,),\[\"(https:|http.*?)\",\d+,\d+\]",
                                                       removed_matched_google_images_thumbnails)


    print('\nFull Resolution Images:')  # in order
    for index, fixed_full_res_image in enumerate(matched_google_full_resolution_images):
        # https://stackoverflow.com/a/4004439/15164646 comment by Frédéric Hamidi
        original_size_img_not_fixed = bytes(fixed_full_res_image, 'ascii').decode('unicode-escape')
        original_size_img = bytes(original_size_img_not_fixed, 'ascii').decode('unicode-escape')
        print(original_size_img)

        # ------------------------------------------------
        # Download original images

        # print(f'Downloading {index} image...')

        opener=urllib.request.build_opener()
        opener.addheaders=[('User-Agent','Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/70.0.3538.102 Safari/537.36 Edge/18.19582')]
        urllib.request.install_opener(opener)

        urllib.request.urlretrieve(original_size_img, f'YOUR_LOCALFOLDER/YOUR_FILE_NAME.jpg')


get_google_images_data()
Enter fullscreen mode Exit fullscreen mode

If you wonder, here's a few attempts for creating regex pattern to extract original resolution images:

1. \[\d+,\[\d+,\"\w+\",,\[\"(https:|http.*?)\",\d+,\d+\]
2. \[\"(https:|http.*?)\",\d+,\d+\]
3. ['|,,]\[\"(https:|http.*?)\",\d+,\d+\]
4. (?:'|,),\[\"(https|http.*?)\",\d+,\d+\] # final
Enter fullscreen mode Exit fullscreen mode

Output from two functions:

-------------------------------------------------
Google Suggested Search Results
-------------------------------------------------

texture pack
https://www.google.com/search?q=minecraft+shaders+8k+photo&tbm=isch&chips=q:minecraft+shaders+8k+photo,online_chips:texture+pack:5UdWXA5mkNo%3D&hl=en-US&sa=X&ved=2ahUKEwiRy8vcgPnxAhUpu6QKHV9FCLsQ4lYoAHoECAEQEA
q:minecraft+shaders+8k+photo,online_chips:texture+pack:5UdWXA5mkNo%3D
...
Suggested Search Thumbnails:
https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcQPsrr--O2yuSyFg-Al3DN0MyhhuO2RcktFCEFiuzs1RoK4oZvS&usqp=CAU
...

------------------------------------------------------
Google Images Results
------------------------------------------------------

8K RESOLUTION Modded Minecraft Gameplay With Ultra Shaders (Yes Really) -  YouTube
youtube.com
https://www.youtube.com/watch?v=_mR0JBLXRLY
...
Google Image Thumbnails:
https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcR6RRuk2k_wMIMt4hhNAatPgmfrDWvAsXrKC90LeBn4GDoySeBQPruapu7ADCSVyORtU48&usqp=CAU
...
Google Full Resolution Images:
https://i.ytimg.com/vi/_mR0JBLXRLY/maxresdefault.jpg
Enter fullscreen mode Exit fullscreen mode

GIF was sped up by 520%

Note: I was going for a process that felt intuitive at the moment of coding. It could be better regex pattern and fewer lines of code where I divide code into parts with regex, but in terms of speed, the GIF above demonstrates that everything happens pretty quickly in 0:00:00:975229 microseconds.

image


Using Google Images API

SerpApi is a paid API with a free plan.

The difference is that iterating over the structured JSON VS figuring out where the decoded thumbnails and their original resolution is pretty awesome.


import os, urllib.request, json # json for pretty output
from serpapi import GoogleSearch


def get_google_images():
    params = {
      "api_key": os.getenv("API_KEY"),
      "engine": "google",
      "q": "pexels cat",
      "tbm": "isch"
    }

    search = GoogleSearch(params)
    results = search.get_dict()

    # print(json.dumps(results['suggested_searches'], indent=2, ensure_ascii=False))
    print(json.dumps(results['images_results'], indent=2, ensure_ascii=False))

    # -----------------------
    # Downloading images

    for index, image in enumerate(results['images_results']):

        print(f'Downloading {index} image...')

        opener=urllib.request.build_opener()
        opener.addheaders=[('User-Agent','Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/70.0.3538.102 Safari/537.36 Edge/18.19582')]
        urllib.request.install_opener(opener)

        urllib.request.urlretrieve(image['original'], f'SerpApi_Images/original_size_img_{index}.jpg')


get_google_images()


--------------------------
'''
Suggested search results:
[
  {
    "name": "wallpaper",
    "link": "https://www.google.com/search?q=minecraft+shaders+8k+photo&tbm=isch&chips=q:minecraft+shaders+8k+photo,online_chips:wallpaper:M78_F4UxoJw%3D&hl=en-US&sa=X&ved=2ahUKEwibusKPuvjxAhWFEt8KHbN0CBUQ4lYoAHoECAEQEQ",
    "chips": "q:minecraft+shaders+8k+photo,online_chips:wallpaper:M78_F4UxoJw%3D",
    "serpapi_link": "https://serpapi.com/search.json?chips=q%3Aminecraft%2Bshaders%2B8k%2Bphoto%2Conline_chips%3Awallpaper%3AM78_F4UxoJw%253D&device=desktop&engine=google&google_domain=google.com&q=minecraft+shaders+8k+photo&tbm=isch",
    "thumbnail": "https://serpapi.com/searches/60fa52ca477c0ec3f75f0d3b/images/3868309500692ce40237282387fb16587c67c8a9bb635eefe35216c182003a4d.jpeg"
  }
...
]

--------------------------

Image results:
[
  {
    "position": 1,
    "thumbnail": "https://serpapi.com/searches/60fa52ca477c0ec3f75f0d3b/images/07dc65d29a3e1094e9c1551efe12324ee8387d268cf2eec92bf0eaed1550eecb.jpeg",
    "source": "reddit.com",
    "title": "8k Minecraft + Shaders: Minecraft",
    "link": "https://www.reddit.com/r/Minecraft/comments/6iamxa/8k_minecraft_shaders/",
    "original": "https://external-preview.redd.it/mAQWN2kUYgFS3fgm6LfYo37AY7i2e_YY8d83_1jTeys.jpg?auto=webp&s=b2bad0e23cbd83426b06e6a547ef32ebbc08e2d2"
  }
...
]
'''
Enter fullscreen mode Exit fullscreen mode

Code in the online IDEGoogle Images API

Outro

If you have any questions or something isn't working correctly or you want to write something else, feel free to drop a comment in the comment section or via Twitter at @serp_api.

Yours,
Dimitry, and the rest of SerpApi Team.

Discussion (0)