TJU-DHD 高分辨率图像数据集

TJU 全称 Tianjin University，DHD 全称 diverse high-resolution dataset，TJU-DHD 是用于目标检测和行人检测的高分辨率数据集，包含 115,354 张高分辨率图像（52% 的图像分辨率为 1624×1200 像素，48% 的图像分辨率至少为 2,560×1,440 像素），709,330 个在形状和外观上存在巨大差异的标记物体。同时，该数据集在季节变化、光照变化和天气变化上也提供丰富多样的变化图像。

TJU stands for Tianjin University, while DHD stands for diverse high-resolution dataset. The TJU-DHD dataset is a high-resolution benchmark for object detection and pedestrian detection. It contains 115,354 high-resolution images, with 52% having a resolution of 1624×1200 pixels and 48% boasting a resolution of at least 2560×1440 pixels. Additionally, the dataset includes 709,330 annotated objects with significant variations in shape and appearance. Furthermore, it features a diverse set of images that exhibit rich variations in seasonal conditions, illumination and weather.

搜集汇总

背景与挑战

背景概述

TJU-DHD是一个用于目标检测和行人检测的高分辨率图像数据集，包含超过11.5万张图像和70.9万个标记物体，图像分辨率多样且覆盖季节、光照及天气变化。该数据集由社区贡献，适用于教育和研究目的。