A vision-language strategy for foundational UI understanding – Google AI Weblog

The computational understanding of person interfaces (UI) is a key step in the direction of reaching clever UI behaviors. Beforehand, we investigated varied UI modeling duties, together with widget captioning, display screen summarization, and command grounding, that handle various interplay eventualities akin to automation and accessibility. We additionally demonstrated how machine studying may help person expertise practitioners enhance UI high quality by diagnosing tappability confusion and offering insights for enhancing UI design. These works together with these developed by others within the discipline have showcased how deep neural networks can probably rework finish person experiences and the interplay design apply.

With these successes in addressing particular person UI duties, a pure query is whether or not we are able to acquire foundational understandings of UIs that may profit particular UI duties. As our first try and reply this query, we developed a multi-task mannequin to handle a spread of UI duties concurrently. Though the work made some progress, just a few challenges stay. Earlier UI fashions closely depend on UI view hierarchies — i.e., the construction or metadata of a cellular UI display screen just like the Doc Object Mannequin for a webpage — that enable a mannequin to straight purchase detailed data of UI objects on the display screen (e.g., their varieties, textual content content material and positions). This metadata has given earlier fashions benefits over their vision-only counterparts. Nevertheless, view hierarchies usually are not at all times accessible, and are sometimes corrupted with lacking object descriptions or misaligned construction data. In consequence, regardless of the short-term features from utilizing view hierarchies, it might in the end hamper the mannequin efficiency and applicability. As well as, earlier fashions needed to cope with heterogeneous data throughout datasets and UI duties, which frequently resulted in advanced mannequin architectures that had been troublesome to scale or generalize throughout duties.

In “Highlight: Cellular UI Understanding utilizing Imaginative and prescient-Language Fashions with a Focus”, accepted for publication at ICLR 2023, we current a vision-only strategy that goals to attain common UI understanding utterly from uncooked pixels. We introduce a unified strategy to characterize various UI duties, the data for which could be universally represented by two core modalities: imaginative and prescient and language. The imaginative and prescient modality captures what an individual would see from a UI display screen, and the language modality could be pure language or any token sequences associated to the duty. We display that Highlight considerably improves accuracy on a spread of UI duties, together with widget captioning, display screen summarization, command grounding and tappability prediction.

Highlight Mannequin

The Highlight mannequin enter features a tuple of three objects: the screenshot, the area of curiosity on the display screen, and the textual content description of the duty. The output is a textual content description or response concerning the area of curiosity. This straightforward enter and output illustration of the mannequin is expressive to seize varied UI duties and permits scalable mannequin architectures. This mannequin design permits a spectrum of studying methods and setups, from task-specific fine-tuning, to multi-task studying and to few-shot studying. The Highlight mannequin, as illustrated within the above determine, leverages present structure constructing blocks akin to ViT and T5 which might be pre-trained within the high-resourced, common vision-language area, which permits us to construct on high of the success of those common area fashions.

As a result of UI duties are sometimes involved with a selected object or space on the display screen, which requires a mannequin to have the ability to give attention to the thing or space of curiosity, we introduce a Focus Area Extractor to a vision-language mannequin that permits the mannequin to focus on the area in gentle of the display screen context.

Specifically, we design a Area Summarizer that acquires a latent illustration of a display screen area primarily based on ViT encodings through the use of consideration queries generated from the bounding field of the area (see paper for extra particulars). Particularly, every coordinate (a scalar worth, i.e., the left, high, proper or backside) of the bounding field, denoted as a yellow field on the screenshot, is first embedded through a multilayer perceptron (MLP) as a set of dense vectors, after which fed to a Transformer mannequin alongside their coordinate-type embedding. The dense vectors and their corresponding coordinate-type embeddings are shade coded to point their affiliation with every coordinate worth. Coordinate queries then attend to display screen encodings output by ViT through cross consideration, and the ultimate consideration output of the Transformer is used because the area illustration for the downstream decoding by T5.

A goal area on the display screen is summarized through the use of its bounding field to question into display screen encodings from ViT through attentional mechanisms.

Outcomes

We pre-train the Highlight mannequin utilizing two unlabeled datasets (an inside dataset primarily based on C4 corpus and an inside cellular dataset) with 2.5 million cellular UI screens and 80 million internet pages. We then individually fine-tune the pre-trained mannequin for every of the 4 downstream duties (captioning, summarization, grounding, and tappability). For widget captioning and display screen summarization duties, we report CIDEr scores, which measure how comparable a mannequin textual content description is to a set of references created by human raters. For command grounding, we report accuracy that measures the share of instances the mannequin efficiently locates a goal object in response to a person command. For tappability prediction, we report F1 scores that measure the mannequin’s skill to inform tappable objects from untappable ones.

On this experiment, we examine Highlight with a number of benchmark fashions. Widget Caption makes use of view hierarchy and the picture of every UI object to generate a textual content description for the thing. Equally, Screen2Words makes use of view hierarchy and the screenshot in addition to auxiliary options (e.g., app description) to generate a abstract for the display screen. In the identical vein, VUT combines screenshots and look at hierarchies for performing a number of duties. Lastly, the unique Tappability mannequin leverages object metadata from view hierarchy and the screenshot to foretell object tappability. Taperception, a follow-up mannequin of Tappability, makes use of a vision-only tappability prediction strategy. We look at two Highlight mannequin variants with respect to the dimensions of its ViT constructing block, together with B/16 and L/16. Highlight drastically exceeded the state-of-the-art throughout 4 UI modeling duties.

We then pursue a tougher setup the place we ask the mannequin to be taught a number of duties concurrently as a result of a multi-task mannequin can considerably scale back mannequin footprint. As proven within the desk under, the experiments confirmed that our mannequin nonetheless performs competitively.

To grasp how the Area Summarizer permits Highlight to give attention to a goal area and related areas on the display screen, we analyze the consideration weights (which point out the place the mannequin consideration is on the screenshot) for each widget captioning and display screen summarization duties. Within the determine under, for the widget captioning process, the mannequin predicts “choose Chelsea group” for the checkbox on the left aspect, highlighted with a purple bounding field. We will see from its consideration heatmap (which illustrates the distribution of consideration weights) on the fitting that the mannequin learns to take care of not solely the goal area of the examine field, but additionally the textual content “Chelsea” on the far left to generate the caption. For the display screen summarization instance, the mannequin predicts “web page displaying the tutorial of a studying app” given the screenshot on the left. On this instance, the goal area is the complete display screen, and the mannequin learns to take care of essential components on the display screen for summarization.

For the widget captioning process, the eye heatmap exhibits the mannequin attending to the checkbox, i.e., the goal object, and the textual content label on its left when producing a caption for the thing. The purple bounding field within the determine is for illustration functions.

For the display screen summarization process that the goal area encloses the complete display screen, the eye heatmap exhibits the mannequin attending to varied places on the display screen that contribute to producing the abstract.

Conclusion

We display that Highlight outperforms earlier strategies that use each screenshots and look at hierarchies because the enter, and establishes state-of-the-art outcomes on a number of consultant UI duties. These duties vary from accessibility, automation to interplay design and analysis. Our vision-only strategy for cellular UI understanding alleviates the necessity to use view hierarchy, permits the structure to simply scale and advantages from the success of huge vision-language fashions pre-trained for the overall area. In comparison with latest massive vision-language mannequin efforts akin to Flamingo and PaLI, Highlight is comparatively small and our experiments present the pattern that bigger fashions yield higher efficiency. Highlight could be simply utilized to extra UI duties and probably advance the fronts of many interplay and person expertise duties.

Acknowledgment

We thank Mandar Joshi and Tao Li for his or her assist in processing the net pre-training dataset, and Chin-Yi Cheng and Forrest Huang for his or her suggestions for proofreading the paper. Due to Tom Small for his assist in creating animated figures on this publish.