Update README.md

README.md

@@ -9,6 +9,9 @@ task_ids: []
 pretty_name: ShowUI_Web
 tags:
 - fiftyone
+- visual-agents
+- gui-grounding
+- os-agents
 - image
 - object-detection
 dataset_summary: '
@@ -46,7 +49,7 @@ dataset_summary: '
 
   # Note: other available arguments include ''max_samples'', etc
 
-  dataset = load_from_hub("harpreetsahota/ShowUI_Web")
+  dataset = load_from_hub("Voxel51/ShowUI_Web")
 
 
   # Launch the App
@@ -60,10 +63,8 @@ dataset_summary: '
 
 # Dataset Card for ShowUI_Web
 
-<!-- Provide a quick summary of the dataset. -->
-
-
 
+![image/png](showui_web.gif)
 
 
 This is a [FiftyOne](https://github.com/voxel51/fiftyone) dataset with 21988 samples.
@@ -84,141 +85,121 @@ from fiftyone.utils.huggingface import load_from_hub
 
 # Load the dataset
 # Note: other available arguments include 'max_samples', etc
-dataset = load_from_hub("harpreetsahota/ShowUI_Web")
+dataset = load_from_hub("Voxel51/ShowUI_Web")
 
 # Launch the App
 session = fo.launch_app(dataset)
 ```
 
+# Web Dataset from ShowUI
 
 ## Dataset Details
 
 ### Dataset Description
 
-<!-- Provide a longer summary of what this dataset is. -->
-
+The Web dataset is a custom-collected corpus of web interface screenshots and element annotations created specifically for training GUI visual agents. It focuses on visually rich UI elements across 22 representative website scenarios (including Airbnb, Booking, AMD, Apple, etc.), purposefully filtering out static text elements to concentrate on interactive components like buttons and checkboxes. This curation strategy was based on the observation that most Vision-Language Models already possess strong OCR capabilities, making visually interactive elements more valuable for training.
 
-
-- **Curated by:** [More Information Needed]
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
+- **Curated by:** Show Lab, National University of Singapore and Microsoft
 - **Language(s) (NLP):** en
-- **License:** [More Information Needed]
-
-### Dataset Sources [optional]
+- **License:** Apache-2.0
 
-<!-- Provide the basic links for the dataset. -->
+### Dataset Sources
 
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
+- **Repository:** https://github.com/showlab/ShowUI (main project repository) and https://huggingface.co/datasets/showlab/ShowUI-web
+- **Paper:** https://arxiv.org/abs/2411.17465
 
 ## Uses
 
-<!-- Address questions around how the dataset is intended to be used. -->
-
 ### Direct Use
 
-<!-- This section describes suitable use cases for the dataset. -->
-
-[More Information Needed]
+The dataset is designed for training vision-language-action models for GUI visual agents operating in web environments. It can be used for:
+- Training models to identify interactive UI elements visually
+- Web element grounding (mapping textual queries to visual elements)
+- Supporting web navigation tasks by providing high-quality visual element references
+- Learning to distinguish between different types of web UI components
 
 ### Out-of-Scope Use
 
-<!-- This section addresses misuse, malicious use, and uses that the dataset will not work well for. -->
-
-[More Information Needed]
+While not explicitly stated in the paper, this dataset would likely be unsuitable for:
+- Training models for desktop or mobile interfaces exclusively
+- General image understanding unrelated to UI navigation
+- Training models where text-based element identification is the primary goal
+- Applications requiring user data or personalized interfaces
 
 ## Dataset Structure
 
-<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->
+The dataset contains:
+- 22,000 web screenshots across 22 representative website scenarios
+- 576,000 annotated UI elements (filtered from an original 926,000)
+- An average of 26 elements per screenshot
+- Element annotations focus on interactive visual components (buttons, checkboxes, etc.)
+- Annotations exclude static text elements (which comprised about 40% of the original data)
+
+## FiftyOne Dataset Structure
+
+**Basic Info:** 21,988 web UI screenshots with interaction annotations
+
+**Core Fields:**
+- `instructions`: ListField(StringField) - List of potential text instructions or UI element texts
+- `detections`: EmbeddedDocumentField(Detections) containing multiple Detection objects:
+  - `label`: Element type (e.g., "ListItem")
+  - `bounding_box`: A list of relative bounding box coordinates in [0, 1] in the following format: `[<top-left-x>, <top-left-y>, <width>, <height>]`
+  - `text`: Text content of element
+- `keypoints`: EmbeddedDocumentField(Keypoints) containing interaction points:
+  - `label`: Element type (e.g., "ListItem")
+  - `points`:  A list of `(x, y)` keypoints in `[0, 1] x [0, 1]`
+  - `text`: Text content associated with the interaction point
 
-[More Information Needed]
+The dataset captures web interface elements and interaction points with detailed text annotations for web interaction research. Each element has both its bounding box coordinates and a corresponding interaction point, allowing for both element detection and precise interaction targeting.
 
 ## Dataset Creation
 
 ### Curation Rationale
 
-<!-- Motivation for the creation of this dataset. -->
-
-[More Information Needed]
+The authors identified that most existing web datasets contain a high proportion of static text elements (around 40%) that provide limited value for training visual GUI agents, since modern Vision-Language Models already possess strong OCR capabilities. Instead, they focused on collecting visually distinctive interactive elements that would better enable models to learn UI navigation skills. This selective approach prioritizes quality and relevance over raw quantity.
 
 ### Source Data
 
-<!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). -->
-
 #### Data Collection and Processing
 
-<!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->
-
-[More Information Needed]
+The authors:
+1. Developed a custom parser using PyAutoGUI [34]
+2. Selected 22 representative website scenarios (including Airbnb, Booking, AMD, Apple, etc.)
+3. Collected multiple screenshots per scenario to maximize annotation coverage
+4. Initially gathered 926,000 element annotations across 22,000 screenshots
+5. Filtered out elements classified as static text, retaining 576,000 visually interactive elements
+6. Focused on elements tagged with categories like "Button" or "Checkbox"
 
 #### Who are the source data producers?
 
-<!-- This section describes the people or systems who originally created the data. It should also include self-reported demographic or identity information for the source data creators if this information is available. -->
-
-[More Information Needed]
-
-### Annotations [optional]
-
-<!-- If the dataset contains annotations which are not part of the initial data collection, use this section to describe them. -->
-
-#### Annotation process
-
-<!-- This section describes the annotation process such as annotation tools used in the process, the amount of data annotated, annotation guidelines provided to the annotators, interannotator statistics, annotation validation, etc. -->
-
-[More Information Needed]
-
-#### Who are the annotators?
-
-<!-- This section describes the people or systems who created the annotations. -->
-
-[More Information Needed]
-
-#### Personal and Sensitive Information
-
-<!-- State whether the dataset contains data that might be considered personal, sensitive, or private (e.g., data that reveals addresses, uniquely identifiable names or aliases, racial or ethnic origins, sexual orientations, religious beliefs, political opinions, financial or health data, etc.). If efforts were made to anonymize the data, describe the anonymization process. -->
-
-[More Information Needed]
+The data was collected from 22 publicly accessible websites across various domains (e-commerce, technology, travel, etc.). The specific screenshots and annotations were produced by the authors of the ShowUI paper (Show Lab, National University of Singapore and Microsoft).
 
 ## Bias, Risks, and Limitations
 
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
-
-[More Information Needed]
+The paper doesn't explicitly discuss biases or limitations specific to this dataset, but potential limitations might include:
+- Limited to 22 website scenarios, which may not represent the full diversity of web interfaces
+- Filtering out static text could limit the model's ability to handle text-heavy interfaces
+- Potential overrepresentation of popular or mainstream websites compared to niche or specialized interfaces
+- May not capture the full range of web accessibility features or alternative UI designs
 
 ### Recommendations
 
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
-
-Users should be made aware of the risks, biases and limitations of the dataset. More information needed for further recommendations.
-
-## Citation [optional]
+Users should be aware that this dataset deliberately excludes static text elements, which makes it complementary to text-focused datasets but potentially incomplete on its own. For comprehensive web navigation models, it should be used alongside datasets that include text recognition capabilities. Additionally, researchers may want to evaluate whether the 22 selected website scenarios adequately represent their target application domains.
 
-<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->
+## Citation
 
 **BibTeX:**
-
-[More Information Needed]
+```bibtex
+@misc{lin2024showui,
+      title={ShowUI: One Vision-Language-Action Model for GUI Visual Agent}, 
+      author={Kevin Qinghong Lin and Linjie Li and Difei Gao and Zhengyuan Yang and Shiwei Wu and Zechen Bai and Weixian Lei and Lijuan Wang and Mike Zheng Shou},
+      year={2024},
+      eprint={2411.17465},
+      archivePrefix={arXiv},
+      primaryClass={cs.CV},
+      url={https://arxiv.org/abs/2411.17465}, 
+}
+```
 
 **APA:**
-
-[More Information Needed]
-
-## Glossary [optional]
-
-<!-- If relevant, include terms and calculations in this section that can help readers understand the dataset or dataset card. -->
-
-[More Information Needed]
-
-## More Information [optional]
-
-[More Information Needed]
-
-## Dataset Card Authors [optional]
-
-[More Information Needed]
-
-## Dataset Card Contact
-
-[More Information Needed]
+Lin, K. Q., Li, L., Gao, D., Yang, Z., Wu, S., Bai, Z., Lei, S. W., Wang, L., & Shou, M. Z. (2024). ShowUI: One Vision-Language-Action Model for GUI Visual Agent. arXiv preprint arXiv:2411.17465.