Add audio-to-audio task category, paper link, and project page link

README.md

@@ -1,19 +1,22 @@
 ---
 license: apache-2.0
+pretty_name: sphere360
+task_categories:
+- audio-to-audio
 tags:
 - audio
 - 360-degree
 - youtube
 - video
 - 360video
-pretty_name: sphere360
 ---
+
 # Sphere360
 
 Sphere360 is a comprehensive dataset of paired 360-degree videos and spatial audio content sourced from YouTube. The collection contains over 103,000 matched 360-degree video and audio clips, representing a total of 288 hours of immersive content. This repository includes both the curated dataset and the essential web crawling and data processing tools used for its compilation.
 
-
-
+The dataset was used to train the model described in the paper [OmniAudio: Generating Spatial Audio from 360-Degree Video](https://huggingface.co/papers/2504.14906).
+Project page: https://OmniAudio-360V2SA.github.io
 
 - [Sphere360](#sphere360)
   - [Copyright](#copyright)
@@ -29,9 +32,6 @@ Sphere360 is a comprehensive dataset of paired 360-degree videos and spatial aud
     - [Data Sources](#data-sources)
     - [Data Cleaning Dependencies](#data-cleaning-dependencies)
 
-
-
-
 ## Copyright
 
 The video data utilized in this study were sourced from the YouTube platform. All content is copyrighted by their respective creators and owners. The videos included in this research adhere to YouTube's terms of service and, where applicable, to Creative Commons licenses. Specifically, videos under the Creative Commons license have been appropriately attributed to the original authors in accordance with the license terms (CC BY 4.0).
@@ -48,8 +48,6 @@ The dataset split configuration can be found in the `dataset/split` directory, c
 - Test set: ~3k samples
 - Each sample duration: 10 seconds
 
-
-
 ## Toolset Environment
 
 #### Python Environment
@@ -81,8 +79,6 @@ This project uses FFmpeg for audio/video data processing. Please configure the [
 
 To use the download scripts provided in this repository, please configure the [yt-dlp](https://github.com/yt-dlp/yt-dlp/tree/master) environment.
 
-
-
 ## Data Crawling
 
 ![DataCrawl](docs/img/crawl.png)
@@ -108,8 +104,6 @@ The general workflow for data crawling is as follows:
 
 **For detailed workflow and script usage, please refer to [docs/crawl.md](docs/crawl.md).**
 
-
-
 ## Data Cleaning
 
 ![DataClean](docs/img/clean.png)
@@ -123,26 +117,22 @@ The cleaning pipeline primarily consists of four dimensions:
 
 **For detailed workflow and script usage, please refer to [docs/clean.md](docs/clean.md).**
 
+## Acknowledgments
 
-
-## Acknowledgments  
-
-This project is built upon the following resources and open-source projects:  
+This project is built upon the following resources and open-source projects:
 
 ### Data Sources
 - **[YouTube Data API v3](https://developers.google.com/youtube/v3/)**
-  
-  Our project utilizes its end-to-end speech recognition model to achieve Voice Detection Filtering. 
 
+  Our project utilizes its end-to-end speech recognition model to achieve Voice Detection Filtering.
 
 ### Data Cleaning Dependencies
-- **[ImageBind](https://github.com/facebookresearch/ImageBind)**  
-  
+- **[ImageBind](https://github.com/facebookresearch/ImageBind)**
+
+Our project employs its cross-modal alignment capability to implement the Audio-Visual Matching Filtering.
+
+- **[SenseVoice](https://github.com/FunAudioLLM/SenseVoice)** (Replace with actual link)
 
-Our project employs its cross-modal alignment capability to implement the Audio-Visual Matching Filtering. 
+  An advanced speech understanding toolkit, licensed under **[License Type]** (e.g., Apache 2.0).
 
-- **[SenseVoice](https://github.com/FunAudioLLM/SenseVoice)** (Replace with actual link)  
-  
-  An advanced speech understanding toolkit, licensed under **[License Type]** (e.g., Apache 2.0).  
-  
   Its end-to-end speech recognition model was instrumental in generating textual metadata for this project.