Datasets:

madebyollin

/

megalith-10m

like 65

Hi!
Thank you for your fantastic work! I'm so impressed!

I want to create my own dataset by using the Flickr API to extract URLs and metadata of all CC-BY licensed content on Flickr. According to the public license information on Flickr's website, it seems that I can generate a dataset of about 95 million items.

I am running the code below, trying to extract URLs in ascending order of the posted date using 'date-posted-asc'. However, the number of results returned by the search is sparse, and I cannot extract the predicted number of URLs.

Which API and args did you use to extract these URLs? (or do you have any plans to make cc-by version of this dataset? haha)

Thank you!

from flickrapi import FlickrAPI, exceptions
import json
import os, time
import time
import zipfile
from tqdm import tqdm

key = "xxxxxxx"
secret = "xxxxxx"
wait_time = 0.
flickr = FlickrAPI(key, secret, format='parsed-json')

per_page = 500
now = time.time() - 60 * 60 * 24  # photos of yesterday

# load checkpoint
with open('meta/count') as f:
    count = int(f.read())
with open('meta/min_upload_date') as f:
    min_upload_date = str(f.read())

while int(min_upload_date) < now:
    try:
        for page in tqdm(range(1, 5)):
            zip_path = f'./data/{count:09n}.zip'
            json_path = f'./{count:09n}.json'
            result = flickr.photos.search(
                page=page,
                per_page=per_page,
                media='photos',
                sort='date-posted-asc',
                safe_search=1,
                license=4,
                media='photos',
                min_upload_date=min_upload_date,
                extras='url_h, url_m, url_s, licence, date_upload, description'
            )
            photos = result['photos']
            json_data = json.dumps(photos)
            with zipfile.ZipFile(zip_path, 'w', zipfile.ZIP_DEFLATED) as zipf:
                zipf.writestr(json_path, json_data)
            count += per_page
            time.sleep(wait_time)
        min_upload_date = photos['photo'][-1]['dateupload']
    except exceptions.FlickrError:
        time.sleep(10)
        continue

    # write checkpoint
    with open('meta/min_upload_date', mode='w') as f:
        f.write(min_upload_date)
    with open('meta/count', mode='w') as f:
        f.write(str(count))