Tencent Video Dataset (TVD)

Tencent Video Dataset (TVD) 是由腾讯、深圳博研科技有限公司和清华大学共同创建的视频数据集，包含86个视频序列，每个序列有65帧，分辨率为4K。数据集内容丰富，涵盖多种场景和动态对象，用于训练神经网络编码工具和测试机器视觉任务，如目标检测和跟踪。创建过程中，视频序列使用高端设备捕捉，并经过专业处理。TVD广泛应用于机器学习相关的研究和标准化活动，特别是在神经网络视频压缩算法的训练中，以及作为视频分析和机器导向任务的测试集。

Tencent Video Dataset (TVD) is a video dataset jointly created by Tencent, Shenzhen Boyan Technology Co., Ltd., and Tsinghua University. It includes 86 video sequences, each containing 65 frames with a 4K resolution. The dataset features rich content covering various scenarios and dynamic objects, and is used for training neural network coding tools and evaluating machine vision tasks such as object detection and tracking. During its creation, the video sequences were captured with high-end equipment and professionally processed. TVD has been widely applied in machine learning-related research and standardization activities, particularly for training neural network-based video compression algorithms, as well as serving as a test set for video analysis and machine-oriented tasks.