-
Notifications
You must be signed in to change notification settings - Fork 9
/
Copy pathindex.html
319 lines (281 loc) · 15.4 KB
/
index.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<!-- Meta tags for social media banners, these should be filled in appropriatly as they are your "business card" -->
<!-- Replace the content tag with appropriate information -->
<meta name="description" content="DESCRIPTION META TAG">
<meta property="og:title" content="ImageInWords"/>
<meta property="og:description" content="ImageInWords: Unlocking Hyper-Detailed Image Descriptions
"/>
<meta property="og:url" content="https://github.com/google/imageinwords"/>
<!-- Path to banner image, should be in the path listed below. Optimal dimenssions are 1200X630-->
<meta property="og:image" content="static/images/Abstract/1.svg" />
<meta property="og:image:width" content="1200"/>
<meta property="og:image:height" content="630"/>
<meta name="twitter:title" content="ImageInWords">
<meta name="twitter:description" content="ImageInWords: Unlocking Hyper-Detailed Image Descriptions">
<!-- Path to banner image, should be in the path listed below. Optimal dimenssions are 1200X600-->
<meta name="twitter:image" content="static/images/Abstract/1.svg">
<meta name="twitter:card" content="summary_large_image">
<!-- Keywords for your paper to be indexed by-->
<meta name="keywords" content="image-captions, image-descriptions, image-to-text, text-to-image">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>ImageInWords</title>
<link rel="icon" type="image/x-icon" href="static/images/favicon.ico">
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro"
rel="stylesheet">
<link rel="stylesheet" href="static/css/bulma.min.css">
<link rel="stylesheet" href="static/css/bulma-carousel.min.css">
<link rel="stylesheet" href="static/css/bulma-slider.min.css">
<link rel="stylesheet" href="static/css/fontawesome.all.min.css">
<link rel="stylesheet"
href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
<link rel="stylesheet" href="static/css/index.css">
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script src="https://documentcloud.adobe.com/view-sdk/main.js"></script>
<script defer src="static/js/fontawesome.all.min.js"></script>
<script src="static/js/bulma-carousel.min.js"></script>
<script src="static/js/bulma-slider.min.js"></script>
<script src="static/js/index.js"></script>
</head>
<body>
<!--<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.3.0/css/all.min.css">-->
<section class="hero">
<div class="hero-body">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column has-text-centered">
<h1 class="title is-1 publication-title">ImageInWords: Unlocking Hyper-Detailed Image Descriptions</h1>
<!-- <div class="publication-venue">EMNLP 2024</div> -->
<div class="is-size-5 publication-authors">
<!-- Paper authors -->
<!-- Paper authors -->
<span class="author-block">
<span class="author-block"><a href="https://www.roopalgarg.com/">Roopal Garg<sup>1</sup></a></span>,
<span class="author-block"><a href="https://cs-people.bu.edu/aburns4/">Andrea Burns<sup>1</sup></a></span>,
<span class="author-block"><a href="https://research.google/people/burcu-karagol-ayan/">Burcu Karagol Ayan<sup>1</sup></a></span>,
<span class="author-block"><a href="https://yonatanbitton.github.io/">Yonatan Bitton<sup>2</sup></a></span>,
<span class="author-block"><a href="https://research.google/people/ceslee-montgomery/">Ceslee Montgomery<sup>1</sup></a></span>,
<span class="author-block"><a href="https://yasumasaonoe.github.io/">Yasumasa Onoe<sup>1</sup></a></span>,
<span class="author-block"><a href="https://twitter.com/andrewbunner">Andrew Bunner<sup>1</sup></a></span>,
<span class="author-block"><a href="https://ranjaykrishna.com/">Ranjay Krishna<sup>3</sup></a></span>,
<span class="author-block"><a href="https://research.google/people/jason-baldridge/">Jason Baldridge<sup>2</sup></a></span>,
<span class="author-block"><a href="http://www.radusoricut.com/">Radu Soricut<sup>1</sup></a></span>
</div>
<div class="is-size-5 publication-authors">
<span class="author-block"><sup>1</sup> Google DeepMind, </span>
<span class="author-block"><sup>2</sup> Google Research, </span>
<span class="author-block"><sup>3</sup> University of Washington</span>
</div>
<!-- ArXiv abstract Link -->
<span class="link-block">
<a href="https://arxiv.org/abs/2405.02793" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="ai ai-arxiv"></i>
</span>
<span>arXiv</span>
</a>
</span>
<!-- Github link -->
<span class="link-block">
<a href="https://github.com/google/imageinwords" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>GitHub</span>
</a>
</span>
<!-- Explorer -->
<span class="link-block">
<a href="https://huggingface.co/spaces/google/imageinwords-explorer" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<p style="font-size:20px">🤗</p>
</span>
<span>Explorer</span>
</a>
</span>
<!-- Dataset -->
<span class="link-block">
<a href="https://huggingface.co/datasets/google/imageinwords" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<p style="font-size:20px">🤗</p>
</span>
<span>Dataset</span>
</a>
</span>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- Paper abstract -->
<section class="section hero is-light">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Abstract</h2>
<div class="content has-text-justified">
<p>Despite the longstanding adage "an image is worth a thousand words," creating accurate and hyper-detailed image descriptions for training Vision-Language models remains challenging. Current datasets typically have web-scraped descriptions that are short, low-granularity, and often contain details unrelated to the visual content. As a result, models trained on such data generate descriptions replete with missing information, visual inconsistencies, and hallucinations. To address these issues, we introduce ImageInWords (IIW), a carefully designed human-in-the-loop annotation framework for curating hyper-detailed image descriptions and a new dataset resulting from this process. We validate the framework through evaluations focused on the quality of the dataset and its utility for fine-tuning with considerations for readability, comprehensiveness, specificity, hallucinations, and human-likeness. Our dataset significantly improves across these dimensions compared to recently released datasets (+66%) and GPT-4V outputs (+48%). Furthermore, models fine-tuned with IIW data excel by +31% against prior work along the same human evaluation dimensions. Given our fine-tuned models, we also evaluate text-to-image generation and vision-language reasoning. Our model's descriptions can generate images closest to the original, as judged by both automated and human metrics. We also find our model produces more compositionally rich descriptions, outperforming the best baseline by up to 6% on ARO, SVO-Probes, and Winoground datasets.</p>
</div><h3>
<br>
<img src="static/images/Abstract/1.svg" style="width: 100%; height: 100%"/>
<img src="static/images/Abstract/2.svg" style="width: 100%; height: 100%"/>
</section>
<!-- Image carousel -->
<section class="hero is-small">
<div class="hero-body">
<div class="container">
<div id="results-carousel" class="carousel results-carousel">
<div class="item">
<h2 class="subtitle is-size-3-tablet has-text-weight-bold has-text-centered has-background-info-light mr-0 pt-3 pb-3">
Dataset
</h2>
<h3 class="subtitle is-size-4-tablet has-text-left pr-4 pl-4 pt-3 pb-3">
<p>
IIW hyper-detailed image description dataset is collected through a new Model Seeded, Sequential Human Augmentation paradigm. Our new annotation guidelines and framework result in descriptions with significantly longer length, containing a greater number of nouns, adjectives, adverbs, and verbs compared to those in prior work.
</p>
<p style="text-align:center;">
<br><br>
<img src="static/images/Dataset/1.png" style="width: 100%; height: 100%"/>
</p>
</h3>
</div>
<div class="item">
<h2 class="subtitle is-size-3-tablet has-text-weight-bold has-text-centered has-background-info-light mr-0 pt-3 pb-3">
Human Authored Data Quality
</h2>
<h3 class="subtitle is-size-4-tablet has-text-left pr-4 pl-4 pt-3 pb-3">
<p>
The dataset achieves State-of-the-art (SoTA) results when evaluated on automated and Human Side-by-Side (SxS) metrics. The automatic readability metrics reflect a more verbose style requiring higher levels of education.
ImageInWords human SxS uses five metrics for evaluation: Comprehensiveness, Specificity, Hallucinations, TLDR quality, and Human-Likeness. Compared to prior work, IIW is rated as significantly better on all metrics with respect to DCI and DOCCI annotations from prior work. Below are the SxS results for human authored IIW data.
</p>
<p style="text-align:center;">
<br><br>
<img src="static/images/Exp-Human-Authored-Data-Quality/1.png" style="width: 80%; height: 80%"/>
</p>
</h3>
</div>
<div class="item">
<h2 class="subtitle is-size-3-tablet has-text-weight-bold has-text-centered has-background-info-light mr-0 pt-3 pb-3">
Fine-tuned Model Quality
</h2>
<h3 class="subtitle is-size-4-tablet has-text-left pr-4 pl-4 pt-3 pb-3">
<p>
Below we also report human SxS on descriptions generated by the IIW trained model. IIW descriptions are compared with models trained with DCI, DOCCI, or GPT-4V.
</p>
<p style="text-align:center;">
<br><br>
<img src="static/images/Exp-Finetuned-Model-Quality/1.png" style="width: 100%; height: 100%"/>
</p>
</h3>
</div>
<div class="item">
<h2 class="subtitle is-size-3-tablet has-text-weight-bold has-text-centered has-background-info-light mr-0 pt-3 pb-3">
Reconstructing Images with Descriptions
</h2>
<h3 class="subtitle is-size-4-tablet has-text-left pr-4 pl-4 pt-3 pb-3">
<p>
ImageInWords models are also evaluated in terms of their utility to generate descriptions for downstream tasks such as text-to-image reconstruction. We find that IIW descriptions result in more accurate generated images, as shown by higher mean rank and clip similarity metrics.
</p>
<p style="text-align:center;">
<br><br>
<img src="static/images/Exp-Reconstruction-t2i/1.png" style="width: 100%; height: 100%"/>
</p>
</h3>
</div>
<div class="item">
<h2 class="subtitle is-size-3-tablet has-text-weight-bold has-text-centered has-background-info-light mr-0 pt-3 pb-3">
Compositional Reasoning with IIW Descriptions
</h2>
<h3 class="subtitle is-size-4-tablet has-text-left pr-4 pl-4 pt-3 pb-3">
<p>
Lastly, IIW finetuned model generated descriptions are evaluated for compositional reasoning tasks. IIW is able to improve accuracy on vision-language compositional reasoning benchmarks ARO and Winoground by several points compared to prior work by generating descriptions with finer grained content.
</p>
<p style="text-align:center;">
<br><br>
<img src="static/images/Exp-Compositional-Reasoning/1.png" style="width: 100%; height: 100%"/>
</p>
</h3>
</div>
</div>
</div>
</div>
</section>
<!-- End image carousel -->
<script
type="module"
src="https://gradio.s3-us-west-2.amazonaws.com/4.29.0/gradio.js"
></script>
<section class="hero is-small is-light">
<div class="hero-body">
<div class="container">
<!-- Paper video. -->
<h4 class="subtitle is-size-1-tablet has-text-weight-bold has-text-centered pr-4 pl-4 pt-3 pb-3">
Dataset Viewer
</h4>
<div class="columns is-centered has-text-centered">
<div class="publication-video">
<!-- <gradio-app src="https://nlphuji-whoops-dataset-viewer.hf.space"></gradio-app> -->
<gradio-app src="https://google-imageinwords-explorer.hf.space"></gradio-app>
</div>
</div>
</div>
</div>
</section>
<!-- Download -->
<section class="section hero is-light">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Download</h2>
<div class="content has-text-justified">
<p>We release the <b>IIW-Benchmark Eval Dataset</b>, IIW human-authored descriptions (image and object level annotations) and comparison to prior work (DCI, DOCCI), machine generated enriched versions of the LocNar and XM3600 datasets are open sourced. The statistics below reflect the extent of the data enrichment (e.g., large increase in length and richness in each part of speech).<br><br>
The datasets are released under a <a href="https://creativecommons.org/licenses/by/4.0/" style="color:blue;">CC-BY-4.0</a> license and can be found at <a href="https://github.com/google/imageinwords/" style="color:blue;">GitHub</a> or be downloaded from <a href="https://huggingface.co/datasets/google/imageinwords" style="color:blue;">Hugging Face</a> in a `jsonl` format.
</p>
</div><h3>
<br>
<img src="static/images/Downloads/1.png" style="width: 100%; height: 100%"/>
</section>
<!--BibTex citation -->
<section class="section" id="BibTeX">
<div class="container is-max-desktop content">
<h2 class="title">BibTeX</h2>
<pre>
<code>
@misc{garg2024imageinwords,
title={ImageInWords: Unlocking Hyper-Detailed Image Descriptions},
author={Roopal Garg and Andrea Burns and Burcu Karagol Ayan and Yonatan Bitton and Ceslee Montgomery and Yasumasa Onoe and Andrew Bunner and Ranjay Krishna and Jason Baldridge and Radu Soricut},
year={2024},
eprint={2405.02793},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
</code>
</div>
</section>
<!--End BibTex citation -->
<footer class="footer">
<div class="container">
<div class="columns is-centered">
<div class="column is-8">
<div class="content">
<p style="color:gray;font-size:9.9px;">
This page was built using the <a href="https://github.com/eliahuhorwitz/Academic-project-page-template" target="_blank">Academic Project Page Template</a>.
</p>
</div>
</div>
</div>
</div>
</footer>
<!-- Statcounter tracking code -->
<!-- You can add a tracker to track page visits by creating an account at statcounter.com -->
<!-- End of Statcounter Code -->
</body>
</html>