ntire23_video_colorization

### Data overview
We are using a video dataset with a large diversity of videos collected from YouTube. The dataset is divided into:

- Train data: A total of 200 paired grayscale and color videos.
- Validation data: 15 videos, the input grayscale videos are provided from the beginning of the challenge and are meant for the participants to get online feedback from the validation server.
- Test data: 15 videos, the participants will receive the input grayscale videos when the final evaluation phase starts and the results will be announced after the challenge is over and the winners are decided.

In addition, we provide grayscale video frames of validation/test data to minimize the differences caused by different video decoding methods.

### Data structure
The dataset has the following structure:
- Training videos (color): 001.mkv, 002.mkv, …, 200.mkv
- Training videos (grayscale): 001.mkv, 002.mkv, …, 200.mkv
- Validation videos (color): 001.mkv, 002.mkv, …, 15.mkv
- Validation videos (grayscale): 001.mkv, 002.mkv, …, 15.mkv
- Validation video frames (grayscale): 001/f001.png, 001/f002.png, …, 015/f300.png
- Test videos (color): 001.mkv, 002.mkv, …, 15.mkv
- Test videos (grayscale): 001.mkv, 002.mkv, …, 15.mkv
- Test video frames (grayscale): 001/f001.png, 001/f002.png, …, 015/f250.png

### Data access
By accessing the data the participants implicitly agree with the terms and conditions of the challenge.

Dataset download: https://www.modelscope.cn/datasets/damo/ntire23_video_colorization/files

Tutorials: first click [数据文件], and then locate the data file (eg. val_frames.zip) and click [下载] to download.

Training data:
- Color video: train_gt.zip
- Grayscale video: train.zip
- Video information: https://www.modelscope.cn/datasets/damo/ntire23_video_colorization/file/view/master/train.csv

Validation data:
- Grayscale video (recommanded for preview): val.zip
- Grayscale video frames (recommanded for inference): val_frames.zip
- Video information: https://www.modelscope.cn/datasets/damo/ntire23_video_colorization/file/view/master/val.csv

Test data:
- Grayscale video (recommanded for preview): (TBU)
- Grayscale video frames (recommanded for inference): (TBU)
- Video information: (TBU)

An example submission for validation: val_eg.zip

### Clone with HTTP
```bash
git clone https://www.modelscope.cn/datasets/damo/ntire23_video_colorization.git
```

### 数据集概览
本数据集收录自YouTube平台，涵盖多样化的视频内容，整体划分为以下三类：
- 训练数据：共计200组灰度与彩色配对视频。
- 验证数据：共15段视频。挑战赛初期将向参赛者提供输入灰度视频，用于参赛者通过验证服务器获取线上反馈。
- 测试数据：共15段视频。在最终评估阶段启动时，参赛者将获得输入灰度视频，所有结果将在挑战赛结束并决出优胜者后公布。

此外，为降低不同视频解码方式带来的差异，我们额外提供了验证集与测试集的灰度视频帧。

### 数据集结构
本数据集的结构如下：
- 训练彩色视频：001.mkv、002.mkv、……、200.mkv
- 训练灰度视频：001.mkv、002.mkv、……、200.mkv
- 验证彩色视频：001.mkv、002.mkv、……、15.mkv
- 验证灰度视频：001.mkv、002.mkv、……、15.mkv
- 验证灰度视频帧：001/f001.png、001/f002.png、……、015/f300.png
- 测试彩色视频：001.mkv、002.mkv、……、15.mkv
- 测试灰度视频：001.mkv、002.mkv、……、15.mkv
- 测试灰度视频帧：001/f001.png、001/f002.png、……、015/f250.png

### 数据使用须知
参赛者访问本数据集即视为同意挑战赛的相关条款与条件。

数据集下载链接：https://www.modelscope.cn/datasets/damo/ntire23_video_colorization/files

使用教程：首先点击[数据文件]，随后找到目标数据文件（例如val_frames.zip）并点击[下载]即可获取。

训练数据：
- 彩色视频：train_gt.zip
- 灰度视频：train.zip
- 视频信息表：https://www.modelscope.cn/datasets/damo/ntire23_video_colorization/file/view/master/train.csv

验证数据：
- 灰度视频（推荐用于预览）：val.zip
- 灰度视频帧（推荐用于推理）：val_frames.zip
- 视频信息表：https://www.modelscope.cn/datasets/damo/ntire23_video_colorization/file/view/master/val.csv

测试数据：
- 灰度视频（推荐用于预览）：(待更新，TBU)
- 灰度视频帧（推荐用于推理）：(待更新，TBU)
- 视频信息表：(待更新，TBU)

验证集示例提交包：val_eg.zip

### 通过HTTP协议克隆
bash
git clone https://www.modelscope.cn/datasets/damo/ntire23_video_colorization.git