Data Structure ?
781.25M
README.md
長期的な未来予測のビデオ表現を用いた教師なし学習
シーケンス予測/再構成を評価するためのテストセット Moving MNIST [782Mb]10,000個のシーケンスを含み、各シーケンスの長さは20で、64x64フレーム内で2つの数字が動く様子を表示します。更新されたarxivファイルの結果では、このテストセットを使用して数値を報告しています。未来予測の場合、指標はクロスエントロピー損失で、各シーケンスの最初の10フレームを条件として、最後の10フレームを予測するために使用されます。
No content available at the moment
No content available at the moment
- Share your thoughts
Go share your ideas~~
ALL
Welcome to exchange and share
Your sharing can help others better utilize data.
Data usage instructions:
I. Data Source and Display Explanation:
- 1. The data originates from internet data collection or provided by service providers, and this platform offers users the ability to view and browse datasets.
- 2. This platform serves only as a basic information display for datasets, including but not limited to image, text, video, and audio file types.
- 3. Basic dataset information comes from the original data source or the information provided by the data provider. If there are discrepancies in the dataset description, please refer to the original data source or service provider's address.
II. Ownership Explanation:
- 1. All datasets on this site are copyrighted by their original publishers or data providers.
III. Data Reposting Explanation:
- 1. If you need to repost data from this site, please retain the original data source URL and related copyright notices.
IV. Infringement and Handling Explanation:
- 1. If any data on this site involves infringement, please contact us promptly, and we will arrange for the data to be taken offline.
- 1. The data originates from internet data collection or provided by service providers, and this platform offers users the ability to view and browse datasets.
- 2. This platform serves only as a basic information display for datasets, including but not limited to image, text, video, and audio file types.
- 3. Basic dataset information comes from the original data source or the information provided by the data provider. If there are discrepancies in the dataset description, please refer to the original data source or service provider's address.
- 1. All datasets on this site are copyrighted by their original publishers or data providers.
- 1. If you need to repost data from this site, please retain the original data source URL and related copyright notices.
- 1. If any data on this site involves infringement, please contact us promptly, and we will arrange for the data to be taken offline.
VIP Download(0.24/day)
Download
15 
874
0
0
collect
Share