FVgNET

FVgNET由317场景组成，显示天然和人造的水果和蔬菜，在受控照明条件下在室内拍摄，覆盖400-1000 nm范围。我们使用由无穷大曲线排列的白纸组成的设置来获取图像，这是摄影中用于将对象与背景隔离的配置。我们实现了良好的光谱覆盖，同时最大限度地减少了最终图像中阴影的存在，方法是用头顶白色发光二极管室内照明、配备玻璃扩散器的150 W卤素灯 (来自Thorlabs的OSL2) 和安装在漫反射反射器中的100 W钨丝灯泡照亮物体。

FVgNET comprises 317 scenes depicting natural and artificial fruits and vegetables, captured indoors under controlled lighting conditions, with a spectral coverage spanning 400–1000 nm. The images were acquired using a setup composed of white papers arranged along an infinity curve, a configuration widely used in photography to isolate subjects from the background. We achieved robust spectral coverage while minimizing shadow artifacts in the final images by illuminating the objects with overhead white LED indoor lighting, a 150 W halogen lamp fitted with a glass diffuser (OSL2 from Thorlabs), and a 100 W tungsten filament lamp mounted within a diffuse reflector.

搜集汇总

背景与挑战

背景概述

FVgNET是一个包含317个场景的数据集，展示了在受控照明条件下拍摄的天然和人造水果和蔬菜图像，覆盖400-1000 nm的光谱范围。该数据集由King Abdullah University of Science and Technology于2022年发布，主要用于计算机视觉和光谱成像研究。