Display consumer interfaces (UIs) and infographics, akin to charts, diagrams and tables, play essential roles in human communication and human-machine interplay as they facilitate wealthy and interactive consumer experiences. UIs and infographics share related design rules and visible language (e.g., icons and layouts), that supply a chance to construct a single mannequin that may perceive, purpose, and work together with these interfaces. Nevertheless, due to their complexity and different presentation codecs, infographics and UIs current a singular modeling problem.
To that finish, we introduce “ScreenAI: A Vision-Language Model for UI and Infographics Understanding”. ScreenAI improves upon the PaLI architecture with the versatile patching technique from pix2struct. We prepare ScreenAI on a singular combination of datasets and duties, together with a novel Display Annotation activity that requires the mannequin to determine UI aspect info (i.e., sort, location and outline) on a display. These textual content annotations present giant language fashions (LLMs) with display descriptions, enabling them to robotically generate question-answering (QA), UI navigation, and summarization coaching datasets at scale. At solely 5B parameters, ScreenAI achieves state-of-the-art outcomes on UI- and infographic-based duties (WebSRC and MoTIF), and best-in-class efficiency on Chart QA, DocVQA, and InfographicVQA in comparison with fashions of comparable measurement. We’re additionally releasing three new datasets: Screen Annotation to judge the structure understanding functionality of the mannequin, in addition to ScreenQA Short and Complex ScreenQA for a extra complete analysis of its QA functionality.
ScreenAI
ScreenAI’s structure relies on PaLI, composed of a multimodal encoder block and an autoregressive decoder. The PaLI encoder makes use of a vision transformer (ViT) that creates picture embeddings and a multimodal encoder that takes the concatenation of the picture and textual content embeddings as enter. This versatile structure permits ScreenAI to unravel imaginative and prescient duties that may be recast as textual content+image-to-text issues.
On prime of the PaLI structure, we make use of a versatile patching technique launched in pix2struct. As an alternative of utilizing a fixed-grid sample, the grid dimensions are chosen such that they protect the native side ratio of the enter picture. This permits ScreenAI to work nicely throughout pictures of varied side ratios.
The ScreenAI mannequin is educated in two phases: a pre-training stage adopted by a fine-tuning stage. First, self-supervised studying is utilized to robotically generate information labels, that are then used to coach ViT and the language mannequin. ViT is frozen throughout the fine-tuning stage, the place most information used is manually labeled by human raters.
 |
| ScreenAI mannequin structure. |
Information era
To create a pre-training dataset for ScreenAI, we first compile an intensive assortment of screenshots from numerous gadgets, together with desktops, cell, and tablets. That is achieved by utilizing publicly accessible web pages and following the programmatic exploration method used for the RICO dataset for cell apps. We then apply a structure annotator, primarily based on the DETR mannequin, that identifies and labels a variety of UI parts (e.g., picture, pictogram, button, textual content) and their spatial relationships. Pictograms endure additional evaluation utilizing an icon classifier able to distinguishing 77 completely different icon sorts. This detailed classification is crucial for decoding the delicate info conveyed via icons. For icons that aren’t lined by the classifier, and for infographics and pictures, we use the PaLI picture captioning mannequin to generate descriptive captions that present contextual info. We additionally apply an optical character recognition (OCR) engine to extract and annotate textual content material on display. We mix the OCR textual content with the earlier annotations to create an in depth description of every display.
 |
A cell app screenshot with generated annotations that embrace UI parts and their descriptions, e.g., TEXT parts additionally include the textual content content material from OCR, IMAGE parts include picture captions, LIST_ITEMs include all their youngster parts. |
LLM-based information era
We improve the pre-training information’s range utilizing PaLM 2 to generate input-output pairs in a two-step course of. First, display annotations are generated utilizing the method outlined above, then we craft a immediate round this schema for the LLM to create artificial information. This course of requires immediate engineering and iterative refinement to search out an efficient immediate. We assess the generated information’s high quality via human validation in opposition to a top quality threshold.
You solely communicate JSON. Don't write textual content that isn’t JSON.
You're given the next cell screenshot, described in phrases. Are you able to generate 5 questions concerning the content material of the screenshot in addition to the corresponding quick solutions to them?
The reply must be as quick as doable, containing solely the mandatory info. Your reply must be structured as follows:
questions: [
{{question: the question,
answer: the answer
}},
...
]
{THE SCREEN SCHEMA}
| A pattern immediate for QA information era. |
By combining the pure language capabilities of LLMs with a structured schema, we simulate a variety of consumer interactions and situations to generate artificial, sensible duties. Specifically, we generate three classes of duties:
- Query answering: The mannequin is requested to reply questions concerning the content material of the screenshots, e.g., “When does the restaurant open?”
- Display navigation: The mannequin is requested to transform a pure language utterance into an executable motion on a display, e.g., “Click on the search button.”
- Display summarization: The mannequin is requested to summarize the display content material in a single or two sentences.
 |
| Block diagram of our workflow for producing information for QA, summarization and navigation duties utilizing present ScreenAI fashions and LLMs. Every activity makes use of a customized immediate to emphasise desired features, like questions associated to counting, involving reasoning, and so on. |
 |
| LLM-generated information. Examples for display QA, navigation and summarization. For navigation, the motion bounding field is displayed in pink on the screenshot. |
Experiments and outcomes
As beforehand talked about, ScreenAI is educated in two phases: pre-training and fine-tuning. Pre-training information labels are obtained utilizing self-supervised studying and fine-tuning information labels comes from human raters.
We fine-tune ScreenAI utilizing public QA, summarization, and navigation datasets and quite a lot of duties associated to UIs. For QA, we use nicely established benchmarks within the multimodal and doc understanding subject, akin to ChartQA, DocVQA, Multi page DocVQA, InfographicVQA, OCR VQA, Web SRC and ScreenQA. For navigation, datasets used embrace Referring Expressions, MoTIF, Mug, and Android in the Wild. Lastly, we use Screen2Words for display summarization and Widget Captioning for describing particular UI parts. Together with the fine-tuning datasets, we consider the fine-tuned ScreenAI mannequin utilizing three novel benchmarks:
- Display Annotation: Allows the analysis mannequin structure annotations and spatial understanding capabilities.
- ScreenQA Quick: A variation of ScreenQA, the place its floor fact solutions have been shortened to include solely the related info that higher aligns with different QA duties.
- Advanced ScreenQA: Enhances ScreenQA Quick with tougher questions (counting, arithmetic, comparability, and non-answerable questions) and incorporates screens with numerous side ratios.
The fine-tuned ScreenAI mannequin achieves state-of-the-art outcomes on numerous UI and infographic-based duties (WebSRC and MoTIF) and best-in-class efficiency on Chart QA, DocVQA, and InfographicVQA in comparison with fashions of comparable measurement. ScreenAI achieves aggressive efficiency on Screen2Words and OCR-VQA. Moreover, we report outcomes on the brand new benchmark datasets launched to function a baseline for additional analysis.
 |
| Evaluating mannequin efficiency of ScreenAI with state-of-the-art (SOTA) fashions of comparable measurement. |
Subsequent, we study ScreenAI’s scaling capabilities and observe that throughout all duties, rising the mannequin measurement improves performances and the enhancements haven’t saturated on the largest measurement.
 |
| Mannequin efficiency will increase with measurement, and the efficiency has not saturated even on the largest measurement of 5B params. |
Conclusion
We introduce the ScreenAI mannequin together with a unified illustration that allows us to develop self-supervised studying duties leveraging information from all these domains. We additionally illustrate the influence of information era utilizing LLMs and examine bettering mannequin efficiency on particular features with modifying the coaching combination. We apply all of those methods to construct multi-task educated fashions that carry out competitively with state-of-the-art approaches on various public benchmarks. Nevertheless, we additionally be aware that our method nonetheless lags behind giant fashions and additional analysis is required to bridge this hole.
Acknowledgements
This undertaking is the results of joint work with Maria Wang, Fedir Zubach, Hassan Mansoor, Vincent Etter, Victor Carbune, Jason Lin, Jindong Chen and Abhanshu Sharma. We thank Fangyu Liu, Xi Chen, Efi Kokiopoulou, Jesse Berent, Gabriel Barcik, Lukas Zilka, Oriana Riva, Gang Li,Yang Li, Radu Soricut, and Tania Bedrax-Weiss for his or her insightful suggestions and discussions, together with Rahul Aralikatte, Hao Cheng and Daniel Kim for his or her assist in information preparation. We additionally thank Jay Yagnik, Blaise Aguera y Arcas, Ewa Dominowska, David Petrou, and Matt Sharifi for his or her management, imaginative and prescient and assist. We’re very grateful toTom Small for serving to us create the animation on this submit.
