ISPW

我们在Wild (ISPW) 数据集中引入ISP。该数据集包括来自最近的智能手机相机的成对原始传感器数据和从高端DSLR相机拍摄的sRGB图像。我们的数据集由200捕获的10百万像素图像对组成，从而产生了28,000多个大小为320 × 320的作物，用于训练，验证和测试。我们对苏黎世原始到RGB (ZRR) 数据集 [9] 和我们的ISPW数据集进行了广泛的烧蚀和最先进的实验。我们的方法在很大程度上优于所有以前的方法，在两个数据集上设置了新的最新技术。在图1中提供了与最佳竞争方法的视觉比较。

We introduce ISP in the Wild (ISPW) dataset. This dataset includes paired raw sensor data from modern smartphone cameras and sRGB images captured by high-end DSLR cameras. Our dataset comprises 200 captured 10-megapixel image pairs, yielding over 28,000 320 × 320 cropped patches for training, validation, and testing. We conducted extensive ablation studies and state-of-the-art experiments on both the Zurich Raw-to-RGB (ZRR) dataset [9] and our ISPW dataset. Our approach significantly outperforms all prior methods, setting new state-of-the-art results on both datasets. Visual comparisons with the best competing methods are provided in Figure 1.