facebook/tgve_plus
收藏资源简介:
--- dataset_info: features: - name: idx dtype: int64 - name: task_name dtype: string - name: input_caption dtype: string - name: output_caption dtype: string - name: instruction dtype: string - name: video_path dtype: string splits: - name: train num_bytes: 287914 num_examples: 1418 download_size: 115426 dataset_size: 287914 --- # Dataset Card for the TGVE+ Test Set ## Dataset Description - **Homepage: https://fdd-video-edit.github.io/** - **Paper: https://arxiv.org/abs/2403.09334** ### Dataset Summary We extend the widely used Text Guided Video Editing (TGVE) benchmark with additional editing tasks. The dataset now comprises seven editing tasks in total: four from the original TGVE and three new tasks, namely (i) object removal (Remove), (ii) object addition (Add), and (iii) texture alterations (Texture). The new tasks utilize the same 76 videos from the original TGVE benchmark. Each row in the dataset consists of the instruction, input/output captions, and the relative path of the video in [TGVE](https://drive.google.com/file/d/1D7ZVm66IwlKhS6UINoDgFiFJp_mLIQ0W/view). For more details please see our [paper](https://arxiv.org/abs/2403.09334) and [project page](https://fdd-video-edit.github.io/). We'd like to thank [InstructVid2Vid](https://github.com/amazon-science/instruct-video-to-video) for creating instructions for the original TGVE tasks. ### Licensing Information Licensed with CC-BY-NC 4.0 License available [here](https://creativecommons.org/licenses/by-nc/4.0/legalcode?fbclid=IwAR2SYZjLRywwUMblkWg0LyAxHVVTloIFlvC-ju3BthIYtOM2jpQHgbeXOsM). ### Citation Information ``` @inproceedings{Singer2024VideoEV, title={Video Editing via Factorized Diffusion Distillation}, author={Uriel Singer and Amit Zohar and Yuval Kirstain and Shelly Sheynin and Adam Polyak and Devi Parikh and Yaniv Taigman}, year={2024}, url={https://api.semanticscholar.org/CorpusID:268385300} } ```
dataset_info: 数据集信息: features: - 特征名: idx 数据类型: 64位整数(int64) - 特征名: task_name 数据类型: 字符串(string) - 特征名: input_caption 数据类型: 字符串(string) - 特征名: output_caption 数据类型: 字符串(string) - 特征名: instruction 数据类型: 字符串(string) - 特征名: video_path 数据类型: 字符串(string) splits: - 划分集名称: train(训练集) 字节数: 287914 样本数量: 1418 下载大小: 115426 数据集总大小: 287914 --- # TGVE+测试集数据集卡片 ## 数据集说明 - **主页**: https://fdd-video-edit.github.io/ - **论文**: https://arxiv.org/abs/2403.09334 ### 数据集概述 我们对广为使用的文本引导视频编辑(Text Guided Video Editing, TGVE)基准数据集进行了扩展,新增了多类视频编辑任务。本数据集目前共包含7类编辑任务:其中4类源自原始TGVE基准数据集,剩余3类为新增任务,分别为(i)物体移除(object removal, Remove)、(ii)物体添加(object addition, Add)以及(iii)纹理修改(texture alterations, Texture)。新增任务复用了原始TGVE基准数据集的全部76段视频。 数据集中的每一条样本均包含指令(instruction)、输入字幕(input_caption)与输出字幕(output_caption),以及对应视频在TGVE数据集仓库中的相对路径,TGVE数据集仓库链接为:https://drive.google.com/file/d/1D7ZVm66IwlKhS6UINoDgFiFJp_mLIQ0W/view。如需了解更多细节,请参阅我们的[论文](https://arxiv.org/abs/2403.09334)与[项目主页](https://fdd-video-edit.github.io/)。 在此我们谨向[InstructVid2Vid](https://github.com/amazon-science/instruct-video-to-video)团队致谢,感谢其为原始TGVE任务生成指令文本。 ### 授权信息 本数据集采用知识共享署名-非商业性使用4.0国际许可协议(CC-BY-NC 4.0)进行授权,许可协议全文可参见[此处](https://creativecommons.org/licenses/by-nc/4.0/legalcode?fbclid=IwAR2SYZjLRywwUMblkWg0LyAxHVVTloIFlvC-ju3BthIYtOM2jpQHgbeXOsM)。 ### 引用信息 @inproceedings{Singer2024VideoEV, title={Video Editing via Factorized Diffusion Distillation}, author={Uriel Singer and Amit Zohar and Yuval Kirstain and Shelly Sheynin and Adam Polyak and Devi Parikh and Yaniv Taigman}, year={2024}, url={https://api.semmatics.org/CorpusID:268385300} }
数据集概述
数据集特征
- idx: 整数类型 (int64)
- task_name: 字符串类型 (string)
- input_caption: 字符串类型 (string)
- output_caption: 字符串类型 (string)
- instruction: 字符串类型 (string)
- video_path: 字符串类型 (string)
数据集划分
- 训练集:
- 数据量: 287914 字节
- 样本数: 1418
数据集大小
- 下载大小: 115426 字节
- 数据集大小: 287914 字节
数据集内容
- 该数据集扩展了原有的Text Guided Video Editing (TGVE) 基准,新增了三个编辑任务:对象移除(Remove)、对象添加(Add)和纹理改变(Texture)。
- 数据集包含76个视频,每个视频对应一条记录,包含指令、输入/输出字幕及视频的相对路径。




