image_url
stringlengths
20
1.79k
caption
stringlengths
14
302
http://lh6.ggpht.com/-IvRtNLNcG8o/TpFyrudaT6I/AAAAAAAAM6o/_11MuAAKalQ/IMG_3422.JPG?imgmax=800
a very typical bus station
http://78.media.tumblr.com/3b133294bdc7c7784b781b45eb9af7be/tumblr_nbirmjpEme1tkk0fco1_500.jpg
sierra looked stunning in this top and this skirt while performing with person at their former university
https://media.gettyimages.com/photos/young-confused-girl-standing-in-front-of-a-wardrobe-picture-id511063329?s=612x612
young confused girl standing in front of a wardrobe
https://thumb1.shutterst…se-125938838.jpg
interior design of modern living room with fireplace in a new house
https://thumb1.shutterst…-a-177023534.jpg
cybernetic scene isolated on white background .
https://media.gettyimages.com/photos/jayz-attends-the-chicago-bulls-vs-brooklyn-nets-playoff-game-at-the-picture-id167112882
gangsta rap artist attends sports team vs playoff game in the borough .
https://prismpub.com/wp-…k_graphic_13.jpg
the jetty : different types of plants to establish a variety of ecosystems .
https://thumb1.shutterst…et-201063638.jpg
traditional ornamental floral paisley bandanna .
https://media.gettyimages.com/photos/bryan-mccabe-of-the-toronto-maple-leafs-skates-against-the-new-jersey-picture-id73480619?k=6&m=73480619&s=612x612&w=0&h=99aa-OK9NQ_tJCllsuA5vifiFfxD-EW13TNYlg9m9D8=
# of the sports team skates against sports team during their game .
http://www.robinhoodshow.com/clients/17668/8642054_org.jpg
by geographical feature category or in the city - a dome for every environment
http://i.dailymail.co.uk/i/pix/2016/08/10/15/371349E500000578-0-image-a-15_1470837871650.jpg
a flight was traveling when the animal got free on tuesday night
https://www.swissinfo.ch…hweiz-15-jpg.jpg
even though agricultural conditions are not ideal for growing tobacco , there is indigenous production .
http://image.dailyfreeman.com/storyimage/DF/20170505/NEWS/170509808/AR/0/AR-170509808.jpg&maxh=400&maxw=667
us state speaks during a demonstration thursday .
https://media.gettyimages.com/photos/actor-john-cothran-jr-arrives-for-the-premiere-of-the-film-black-in-picture-id539614492?s=612x612
actor arrives for the premiere of the film
http://images.gmanews.tv/webpics/2016/11/640_2_2016_11_29_17_19_15.jpg
celebrities start decorating for the christmas season lifestyle
http://images.slideplayer.com/2/765769/slides/slide_46.jpg
functions of government : 1 . form a more perfect union
https://media.gettyimages.com/photos/actress-kirsten-dunst-attends-the-premiere-of-fxs-fargo-season-2-at-picture-id491859790?s=612x612
actor attends the premiere of season
http://www.bostonherald.com/sites/default/files/styles/gallery/public/media/2016/08/15/081516patsnl15.jpg?itok=ka5cQo6t
american football player on the field during joint training camp .
http://globe-views.com/dcim/dreams/court/court-04.jpg
companies have gone to court for the right to lie
https://ep1.pinkbike.org…/p4pb6973204.jpg
all shots by by person and rider shots can be found on his website .
http://2.bp.blogspot.com/-cZpqM2kr2uc/U6v-pnIxM9I/AAAAAAAAaek/OFq7OBbqkv4/s1600/photo-of-a-deer-with-a-wildfire-in-the-background-hd-deer-wallpapers.jpg
photo of a deer and wildfire
https://media.gettyimages.com/photos/high-angle-view-of-a-businessman-lying-on-a-table-and-singing-picture-id56677061?s=612x612
high angle view of a businessman lying on a table and singing
https://i.pinimg.com/736…ealthy-foods.jpg
this is real fast food !
https://us.123rf.com/450…ground.jpg?ver=6
safe deposit with money around it on a white background photo
https://timedotcom.files…02/dv1633968.jpg
the giraffe before he was shot dead then autopsied in the presence of the zoo 's visitors , despite an online petition to save him signed by thousands of animal lovers
http://www.golfeurope.com/photo-galleries/images/preview/39779.jpg
dunes lay the blueprint for the back nine .
http://l7.alamy.com/zooms/7f4ac7c33e5841b6a94a35915e5ca1d7/portrait-of-a-smiling-woman-stroking-her-dog-lying-on-couch-d9g1bc.jpg
portrait of a smiling woman stroking her dog lying on couch
http://l7.alamy.com/zooms/b7385d563da542c2be31a30be3a069a6/young-business-woman-on-a-bench-g0dktp.jpg
young business woman on a bench
http://img.bleacherreport.net/img/images/photos/003/558/542/hi-res-25da6f4fff0b052f7c72a5af985559ff_crop_north.jpg?h=533&w=800&q=70&crop_x=center&crop_y=top
american football player looks downfield during the second half of a football game against sports team
http://davidbarrie.typepad.com/.a/6a00d834519d9469e2013485c5657d970c-800wi
... and local people to deliver a new bridge
https://media.gettyimages.com/photos/actress-kate-hudson-arrives-to-the-premiere-of-lionsgates-my-best-picture-id82832165?s=612x612
actor arrives to the premiere
http://4.bp.blogspot.com/-NTT96j0rYZ0/U8lsad0ruCI/AAAAAAABBMU/LKmGZp-Hsg8/s1600/funny-animals-117-40.jpg
funny animals of the week , animal pictures
https://www.bailiwickexp…41ac_f587432.jpg
see the inspiring way this woman documented her travels on her prosthetic leg
http://www.brian-coffee-spot.com/wp-content/uploads/wow-slider-plugin/555/images/dsc_0838.jpg
the sign promises as much as the glorious blue sky .
https://media.gettyimages.com/photos/architectural-details-of-a-bridge-picture-id511832249?s=612x612
architectural details of a bridge
http://l7.alamy.com/zooms/a4197bf168734006b10e02a8bebdd34e/people-tour-and-enjoy-the-high-line-public-park-in-new-york-city-during-ga67w4.jpg
people tour and enjoy the public park during summer
https://i.pinimg.com/736…girl-posters.jpg
interesting 1930 's poster for a cosmetic company with stores .
http://cdn.newsapi.com.au/image/v1/af347cd6569987c49a71fe05870753c5?width=1024
racecar driver steers his car during video game subject .
https://us.123rf.com/450…ackground-ve.jpg
vintage elegant floral card with frame decorated with black and white lilies on a pink background .
https://ak4.picdn.net/sh…4824/thumb/1.jpg
heavy snow falls over a snow lined river .
http://homeklondike.org/wp-content/uploads/2015/06/2-Bright-living-room-in-the-attic1.jpg
bright living room in the attic
https://media.gettyimages.com/photos/actor-kevin-mchale-attends-the-3rd-annual-nautica-oceana-beach-house-picture-id472759470?s=612x612
pop artist attends the 3rd annual at guest house
https://thumb1.shutterst…-a-675197230.jpg
illustration of a map , its flag and a comic balloon with a soccer ball in a not allowed signal
https://media.gettyimages.com/photos/musician-sting-performs-on-stage-at-the-46th-annual-grammy-awards-at-picture-id2949702?s=612x612
rock artist performs on stage at awards held
https://us.123rf.com/450…ration.jpg?ver=6
green sea turtle isolated on a white background 3d illustration
http://i.dailymail.co.uk/i/pix/2017/07/27/17/42B4031600000578-0-image-m-9_1501174323232.jpg
person , was surprised by the staff
https://ak6.picdn.net/sh…3756/thumb/1.jpg
red and white flag on the mast
http://ichef.bbci.co.uk/onesport/cps/624/cpsprodpb/13A58/production/_89827408_kiss.jpg
football player celebrates scoring for football team against football team in the final
http://l7.alamy.com/zooms/b365ec4718354a5bab1eb4413bbb94ae/detroit-michigan-the-vw-cross-coup-gte-concept-plug-in-hybrid-car-ef27pw.jpg
concept plug - in hybrid car on display
https://s-media-cache-ak…73a2bb16a46f.jpg
a pencil drawing of a zebra and her baby .
http://file.vintageadbrowser.com/3iodaozwajsgwb.jpg
airline -- reasons why person leads the way in experience
http://www.remonline.com/wp-content/uploads/2017/06/Stone-South-development-1.jpg
ninety per cent of the units have balconies with views .
https://media.gettyimages.com/photos/demonstration-of-a-group-of-people-practicing-their-rights-picture-id172184947?s=612x612
a demonstration of a group of people practicing their rights
http://l7.alamy.com/zooms/e163858450394e5aa21b226debd6d01f/sylvester-stallone-jennifer-flavin-and-daughters-the-expendables-2-dmrdh6.jpg
actor and daughters uk premiere held
https://image.shuttersto…ite-54561763.jpg
a fine , grainy vector pattern in black and white .
https://thumb1.shutterst…nd-761692834.jpg
seamless border of orange roses and paisley , pattern on a white background .
http://l7.alamy.com/zooms/50937c69ef8f4289a71ef77f1ccf593f/ghana-central-region-komenda-students-in-front-of-a-school-cp4tgf.jpg
students in front of a school
https://www.papercitymag…-40-680x1024.jpg
museum , opened is among the 20th - century 's most significant buildings .
http://l7.alamy.com/zooms/faf62ef929df42f49afde816f043a2a6/fraser-river-as-it-meets-the-pacific-ocean-where-delta-meets-richmond-erk04d.jpg
river as it meets bodies of water where airline meets
https://media.istockphoto.com/photos/young-rock-star-jamming-on-a-guitar-picture-id185283450?k=6&m=185283450&s=612x612&w=0&h=ZRLfgSeOLNpmSH9S9uITPMGYuk7_NxoV136NdgiXoJU=
young rock star jamming on a guitar
http://l7.alamy.com/zooms/8e24000f1ff244508864657b048b3a8b/a-moored-fishing-boat-viewed-from-bryher-isles-of-scilly-cornwall-b7g5nt.jpg
a moored fishing boat viewed from island .
https://image.shuttersto…nd-516650239.jpg
vector illustration of person isolated on a white background
http://www.readingchronicle.co.uk/resources/images/6179443.jpg?display=1&htype=80&type=responsive-gallery
emergency services were called after a car smashed through a set of traffic lights
http://c8.alamy.com/comp/K32RAK/sheep-with-a-black-face-fenced-in-on-a-winter-day-K32RAK.jpg
sheep with a black face fenced in on a winter day
http://l7.alamy.com/zooms/1ff618175b9d44a883c2a6bdf5b28fa1/a-small-park-with-american-flags-and-a-yellow-banner-around-a-tree-acfbd8.jpg
a small park with flags and a yellow banner around a tree in support of troops
https://thumb1.shutterst…on-632286614.jpg
isolated water glass on a white background
https://s-media-cache-ak…513a267298ef.jpg
when choosing your new outdoor color palette , opt for something that will enhance your home 's architectural style and give you plenty of curb appeal .
https://cdn0.weddingwire…y-7_4_108303.png
person at a corporate event
https://media.istockphoto.com/photos/bulldozer-on-a-building-site-picture-id467448303?k=6&m=467448303&s=612x612&w=0&h=4HNBM4KKb6ZmLpF98JWA2_eog44rbKMnfKLqR5W92hg=
bulldozer on a building site
http://jesuitpride.com/images/2017/7/11/Fairfield_Prep_Golf_2017_10_1280x801_.jpg
person finishes at the top
https://ak1.picdn.net/sh…9011/thumb/9.jpg
country shape animated on the satellite map of the globe
https://seekingfelicity.…/06/p1130733.jpg
paths are clearly marked with signs like this .
https://i.pinimg.com/736…mpson-rancho.jpg
putts for birdie on the second hole during the final round of the golf tournament .
https://ak1.picdn.net/sh…size(height:160)
person shot from a cliff looking out at the lake and horizon .
https://i.pinimg.com/736…ee-figurines.jpg
sending this angel your way ... it lights the way with it 's tiny candle .
https://i.pinimg.com/736…t-style-muse.jpg
i love the colours of her clothes .
http://www.southampton.ac.uk/~imw/jpg-Kimmeridge/9KM-kimloc-recovered-m.jpg
a very simplified location and geological map of the area and adjacent coast
https://www.carsdir.com/…966-211995-1.jpg
ask owner for photos of the car
https://3c1703fe8d.site.…thestateofth.jpg
the state of the world 's forests
http://oneindiaonepeople.com/wp-content/uploads/2014/02/13.jpg
the mountain slopes covered with powdery snow are popular with visitors , especially foreign tourists
https://ak7.picdn.net/sh…2697/thumb/1.jpg
farm tractor is moving on the field , cultivating land
http://l7.alamy.com/zooms/6ce413683932455f9214da51ef6c8a78/master-playing-with-his-little-golden-retriever-dog-on-the-lawn-bxgte5.jpg
master playing with his little golden retriever dog on the lawn
https://tipsfromashton.f…2016/10/blog.jpg
year later the small shrubs doubled in size .
https://www.thesun.co.uk…?strip=all&w=960
according to the model , she regularly gets told it looks
https://4.bp.blogspot.co…angela-akins.jpg
person and the fiancee at their engagement party .
https://www.birthdaywish…py-and-quote.jpg
good night id been working like a dog i should be sleeping like a log
https://acanadianincambo…ar-day-4-052.jpg
the water was never deeper than your chest .
http://l7.alamy.com/zooms/252ec157a42f4615a4c9639a6b1148f1/twilight-over-palazzo-dei-priori-and-the-medieval-town-of-volterra-hfpkxa.jpg
twilight over italian gothic structure and the medieval town
https://lynismael.com/wp…p_w768_h534).jpg
a simple wedding cake with lego bride and groom topper and cake pops .
https://i.pinimg.com/736…91a79c04a2e0.jpg
simple custom leather pulls for your dresser , cabinets , or doors .
http://www.hollowtop.com/journals/Tongue_River_Photos/Sunny_Morning.jpg
sunny woodland morning along river .
https://i.pinimg.com/736…e-zombie-art.jpg
person , why u turn me on with your designs ? i believe i will bepinning some more of his work here in a bit .
http://media.gettyimages.com/photos/entertainer-beyonce-performs-on-stage-during-the-formation-world-tour-picture-id538724332
pop artist performs on stage .
https://media.istockphoto.com/photos/industrial-plants-in-the-distance-at-night-picture-id479157539?k=6&m=479157539&s=612x612&w=0&h=0EmjeUArhPcd38poc3cqx_bq0-IKvnfh1qhlxCfesM4=
industrial plants in the distance at night
http://l7.alamy.com/zooms/476f507779b646c4b9ae7e3894c76eb8/the-empire-state-building-and-one-world-trade-center-light-up-at-night-fcn345.jpg
building and skyscraper light up at night
https://i.pinimg.com/736…tiful-people.jpg
actor is an actor who started out as a lawyer .
https://i.pinimg.com/736…living-rooms.jpg
color good with green couch in living room ... favorite ... i think maybe a little too peach ?
http://www.fourintravels.com/wp-content/uploads/2013/12/The-villas-from-the-front1.jpg
the villas from the front
https://ak8.picdn.net/sh…1628/thumb/1.jpg
tree branches swing at real time with the wind while clouds move fast in time lapse
https://burpple-1.imgix.net/foods/2cb7a04812e4af46eb01248319_original.?w=645&dpr=1&fit=crop&q=80
my coffee of the day !

Dataset Card for Conceptual Captions

Dataset Summary

Conceptual Captions is a dataset consisting of ~3.3M images annotated with captions. In contrast with the curated style of other image caption annotations, Conceptual Caption images and their raw descriptions are harvested from the web, and therefore represent a wider variety of styles. More precisely, the raw descriptions are harvested from the Alt-text HTML attribute associated with web images. To arrive at the current version of the captions, we have developed an automatic pipeline that extracts, filters, and transforms candidate image/caption pairs, with the goal of achieving a balance of cleanliness, informativeness, fluency, and learnability of the resulting captions.

Dataset Preprocessing

This dataset doesn't download the images locally by default. Instead, it exposes URLs to the images. To fetch the images, use the following code:

from concurrent.futures import ThreadPoolExecutor
from functools import partial
import io
import urllib

import PIL.Image

from datasets import load_dataset
from datasets.utils.file_utils import get_datasets_user_agent


USER_AGENT = get_datasets_user_agent()


def fetch_single_image(image_url, timeout=None, retries=0):
    for _ in range(retries + 1):
        try:
            request = urllib.request.Request(
                image_url,
                data=None,
                headers={"user-agent": USER_AGENT},
            )
            with urllib.request.urlopen(request, timeout=timeout) as req:
                image = PIL.Image.open(io.BytesIO(req.read()))
            break
        except Exception:
            image = None
    return image


def fetch_images(batch, num_threads, timeout=None, retries=0):
    fetch_single_image_with_args = partial(fetch_single_image, timeout=timeout, retries=retries)
    with ThreadPoolExecutor(max_workers=num_threads) as executor:
        batch["image"] = list(executor.map(fetch_single_image_with_args, batch["image_url"]))
    return batch


num_threads = 20
dset = load_dataset("google-research-datasets/conceptual_captions")
dset = dset.map(fetch_images, batched=True, batch_size=100, fn_kwargs={"num_threads": num_threads})

Supported Tasks and Leaderboards

  • image-captioning: This dataset can be used to train model for the Image Captioning task. The leaderboard for this task is available here. Official submission output captions are scored against the reference captions from the hidden test set using this implementation of the CIDEr (primary), ROUGE-L and SPICE metrics.

Languages

All captions are in English.

Dataset Structure

Data Instances

unlabeled

Each instance in this configuration represents a single image with a caption:

{
  'image_url': 'http://lh6.ggpht.com/-IvRtNLNcG8o/TpFyrudaT6I/AAAAAAAAM6o/_11MuAAKalQ/IMG_3422.JPG?imgmax=800',
  'caption': 'a very typical bus station'
}

labeled

Each instance in this configuration represents a single image with a caption with addtional machine-generated image labels and confidence scores:

{
  'image_url': 'https://thumb1.shutterstock.com/display_pic_with_logo/261388/223876810/stock-vector-christmas-tree-on-a-black-background-vector-223876810.jpg',
  'caption': 'christmas tree on a black background .',
  'labels': ['christmas tree', 'christmas decoration', 'font', 'text', 'graphic design', 'illustration','interior design', 'tree', 'christmas eve', 'ornament', 'fir', 'plant', 'pine', 'pine family', 'graphics'],
  'MIDs': ['/m/025nd', '/m/05fc9mj', '/m/03gq5hm', '/m/07s6nbt', '/m/03c31', '/m/01kr8f', '/m/0h8nzzj', '/m/07j7r', '/m/014r1s', '/m/05ykl4', '/m/016x4z', '/m/05s2s', '/m/09t57', '/m/01tfm0', '/m/021sdg'],
  'confidence_scores': [0.9818305373191833, 0.952756941318512, 0.9227379560470581, 0.8524878621101379, 0.7597672343254089, 0.7493422031402588, 0.7332468628883362, 0.6869218349456787, 0.6552258133888245, 0.6357356309890747, 0.5992692708969116, 0.585474967956543, 0.5222904086112976, 0.5113164782524109, 0.5036579966545105]
}

Data Fields

unlabeled

  • image_url: Static URL for downloading the image associated with the post.
  • caption: Textual description of the image.

labeled

  • image_url: Static URL for downloading the image associated with the post.
  • caption: Textual description of the image.
  • labels: A sequence of machine-generated labels obtained using the Google Cloud Vision API.
  • MIDs: A sequence of machine-generated identifiers (MID) corresponding to the label's Google Knowledge Graph entry.
  • confidence_scores: A sequence of confidence scores denoting how likely the corresponing labels are present on the image.

Data Splits

unlabeled

The basic version of the dataset split into Training and Validation splits. The Training split consists of 3,318,333 image-URL/caption pairs and the Validation split consists of 15,840 image-URL/caption pairs.

labeled

The labeled version of the dataset with a single. The entire data is contained in Training split, which is a subset of 2,007,090 image-URL/caption pairs from the Training set of the unlabeled config.

Dataset Creation

Curation Rationale

From the paper:

In this paper, we make contributions to both the data and modeling categories. First, we present a new dataset of caption annotations Conceptual Captions (Fig. 1), which has an order of magnitude more images than the COCO dataset. Conceptual Captions consists of about 3.3M himage, descriptioni pairs. In contrast with the curated style of the COCO images, Conceptual Captions images and their raw descriptions are harvested from the web, and therefore represent a wider variety of styles.

Source Data

Initial Data Collection and Normalization

From the homepage:

For Conceptual Captions, we developed a fully automatic pipeline that extracts, filters, and transforms candidate image/caption pairs, with the goal of achieving a balance of cleanliness, informativeness, fluency, and learnability of the resulting captions. Because no human annotators are involved, the Conceptual Captions dataset generation process is highly scalable.

To generate this dataset, we started with a Flume pipeline that processes billions of Internet webpages, extracting, filtering, and processing candidate image and caption pairs, and keeping those that pass through several filters.

We first screen for certain properties like size, aspect ratio, adult content scores. These filters discard more than 65% of the candidates. Next, we use Alt-Texts for text-based filtering, removing captions with non-descriptive text (such as SEO tags or hashtags); we also discard texts with high sentiment polarity or adult content scores, resulting in just 3% of the incoming candidates passing through.

In the next step, we filter out candidates for which none of the text tokens can be mapped to the visual content of the image. We use image classifiers (e.g., Google Cloud Vision APIs) to assign class labels to images and match these labels against the candidate text (allowing morphological transformations), discarding >around 60% of the candidates that reach this stage.

The candidates passing the above filters tend to be good Alt-text image descriptions. However, a large majority of these use proper names (for people, venues, locations, etc.), brands, dates, quotes, etc. This creates two distinct problems. First, some of these cannot be inferred based on the image pixels alone. This is problematic because unless the image has the necessary visual information it is not useful for training. Second, even if the proper names could be inferred from the image it is extremely difficult for a model to learn to perform both fine-grained classification and natural-language descriptions simultaneously. We posit that if automatic determination of names, locations, brands, etc. is needed, it should be done as a separate task that may leverage image meta-information (e.g. GPS info), or complementary techniques such as OCR.

We address the above problems with the insight that proper names should be replaced by words that represent the same general notion, i.e., by their concept. For example, we remove locations (“Crowd at a concert in Los Angeles“ becomes “Crowd at a concert”), names (e.g., “Former Miss World Priyanka Chopra on the red carpet” becomes “actor on the red carpet”), proper noun modifiers (e.g., “Italian cuisine” becomes just “cuisine”) and noun phrases (e.g., “actor and actor” becomes “actors”). Around 20% of the samples are discarded during this transformation because it can leave sentences too short, or otherwise inconsistent.

Finally, we perform another round of filtering to identify concepts with low-count. We cluster all resolved entities (e.g., “actor”, “dog”, “neighborhood”, etc.) and keep only the candidate types which have a count of over 100 mentions. This retains around 16K entity concepts such as: “person”, “actor”, “artist”, “player” and “illustration”. The less frequent ones that we dropped include “baguette”, “bridle”, “deadline”, “ministry” and “funnel”.

Who are the source language producers?

Not specified.

Annotations

Annotation process

Annotations are extracted jointly with the images using the automatic pipeline.

Who are the annotators?

Not specified.

Personal and Sensitive Information

[More Information Needed]

Considerations for Using the Data

Social Impact of Dataset

[More Information Needed]

Discussion of Biases

[More Information Needed]

Other Known Limitations

[More Information Needed]

Additional Information

Dataset Curators

Piyush Sharma, Nan Ding, Sebastian Goodman and Radu Soricut.

Licensing Information

The dataset may be freely used for any purpose, although acknowledgement of Google LLC ("Google") as the data source would be appreciated. The dataset is provided "AS IS" without any warranty, express or implied. Google disclaims all liability for any damages, direct or indirect, resulting from the use of the dataset.

Citation Information

@inproceedings{sharma2018conceptual,
  title = {Conceptual Captions: A Cleaned, Hypernymed, Image Alt-text Dataset For Automatic Image Captioning},
  author = {Sharma, Piyush and Ding, Nan and Goodman, Sebastian and Soricut, Radu},
  booktitle = {Proceedings of ACL},
  year = {2018},
}

Contributions

Thanks to @abhishekkrthakur and @mariosasko for adding this dataset.

Downloads last month
16,238
Papers with Code

Data Sourcing report

powered
by Spawning.ai

Some elements in this dataset have been identified as opted-out, or opted-in, by their creator.

Opted-out
Opted-in

Models trained or fine-tuned on google-research-datasets/conceptual_captions