TUM Kitchen 厨房活动数据集

TUM Kitchen 是一个动作识别数据集，涵盖厨房的日常活动。该数据集包含由 4 台摄像机拍摄的 20 个视频序列，拍摄视角相互重叠。摄像机以 25 fps 的速度从 4 个视角捕捉场景，每一个 RGB 帧的分辨率为 384×288 像素。数据分别标注为左手、右手和人的躯干。

TUM Kitchen is an action recognition dataset focused on daily kitchen activities. It contains 20 video sequences captured by 4 cameras with mutually overlapping viewing angles. The cameras capture the scene at 25 fps across four perspectives, with each RGB frame having a resolution of 384×288 pixels. Annotations are provided for the left hand, right hand, and human torso respectively.

搜集汇总

背景与挑战

背景概述

TUM Kitchen 厨房活动数据集是一个用于动作识别的数据集，专注于厨房日常活动。它包含20个视频序列，由4台重叠视角的摄像机以25 fps帧率和384×288像素分辨率拍摄，数据标注包括左手、右手和人的躯干。