Derwya/half-life_player_seg
收藏资源简介:
该数据集包含来自《半条命》游戏的截图,标注了玩家模型的多边形掩码。它专门设计用于训练YOLO分割模型,以支持实时玩家追踪和瞄准辅助算法。数据集格式为YOLO分割(包含归一化多边形坐标的txt文件),总图像数为598张(平衡包含空背景帧),分为训练集(539张)和验证集(59张),类别为1类(玩家)。数据集结构遵循标准YOLO格式,包括图像文件夹、标签文件夹和配置文件。用途为计算机视觉在游戏环境中的研究,通过分割掩码而非标准边界框,训练模型可精确定位玩家轮廓,从而提高瞄准算法的精确度,减少复杂背景下的误锁定,并允许追踪器计算真实质心或头部位置。
This dataset contains in-game screenshots from Half-Life, annotated with polygon masks for player models. It is specifically designed to train YOLO segmentation models for real-time player tracking and aim assistance algorithms. The dataset format is YOLO Segmentation (txt files with normalized polygon coordinates), with a total of 598 images (balanced to include empty background frames), split into 539 images for training and 59 for validation, and includes 1 class (player). The structure follows the standard YOLO format, with images/train/, images/val/, labels/train/, labels/val/, and a data.yaml configuration file. It is intended for computer vision research in gaming environments, using segmentation masks to enable precise player outline detection, improving aim algorithm accuracy, reducing false lock-ons on complex backgrounds, and allowing tracker calculations for true center-of-mass or head positions.




