isp-uv-es/IPL-Cityscapes-Grayscale-Illuminants

IPL-Cityscapes-Grayscale-Illuminants数据集是基于著名自动驾驶语义分割数据集Cityscapes的灰度版本，通过修改光照条件生成的变体数据集。具体生成过程包括：首先将图像转换为灰度，然后生成平坦光光谱并计算像素反射率，接着生成不同主波长（颜色）和饱和度的新图像。该方法应用于100张Cityscapes图像，每张原始图像生成120个修改版本（20种颜色和6种饱和度）。数据集结构包括original_images文件夹（包含100张原始灰度图像）和modified_images文件夹（包含100个子文件夹，每个子文件夹对应一张原始图像的120个修改图像）。修改图像命名格式为image_name_X_tono_T_sat_S.png，其中X为图像索引（1到100），T为颜色索引（1到20），S为饱和度索引（1到6）。标签对应IPL-CityscapesIlluminants-dataset中的labels文件夹。完整数据集（原始灰度图像加修改图像）大小为30GB。IPL-Cityscapes-Grayscale-Illuminants dataset is an illuminant modified dataset version of the grayscale famous autonomous driving semantic segmentation Cityscapes dataset. For each image, it is first converted to grayscale. Then, a flat (constant) light spectrum is generated and the pixel reflectances that obtain the RGB pixel values are computed. Once the pixel reflectances are obtained, different light spectrums of different dominant wavelengths (colors) and saturations are generated to compute the new modified images. The method is applied to 100 Cityscapes images, with each original image generating 120 modified versions (20 colors and 6 saturations). The dataset structure includes an original_images folder containing the 100 original grayscale Cityscapes images, and a modified_images folder with 100 subfolders from image_1 to image_100, each containing the 120 modified images for each original image. Modified images are named as image_name_X_tono_T_sat_S.png, where X is the image index from 1 to 100, T is the color index from 1 to 20, and S is the saturation index from 1 to 6. Labels are found in the labels folder of IPL-CityscapesIlluminants-dataset. The complete dataset (original grayscale + modified images) has a size of 30 GB.