Real world blurred image dataset
Download link: http://suo.nz/2nk7w2
Training set: 3,758 image pairs of 182 different scenes. Test set: 980 image pairs of 50 different scenes.
I-HAZE image dehazing dataset
Download link: http://suo.nz/2fNUrl
This dataset contains 35 hazy image pairs and corresponding haze-free (ground truth) indoor images. Unlike most existing dehazing databases, haze images are generated using real haze produced by professional haze machines. To simplify color calibration and improve evaluation of dehazing algorithms, each scene includes a MacBeth Color Inspector. Additionally, since the images are captured in a controlled environment, both haze-free and hazy images are captured under the same lighting conditions.
UIEB Underwater Image Enhancement Dataset
Download link: http://suo.nz/3dY13O
The dataset includes 950 real underwater images, 890 of which have corresponding reference images. And the remaining 60 underwater images for which satisfactory reference images cannot be obtained are regarded as challenging data.
GoPro deblurred dataset
Download link: http://suo.nz/36rNZ7
The GoPro dataset used for deblurring consists of 3,214 blurred images with a size of 1,280×720, divided into 2,103 training images and 1,111 testing images. This dataset consists of pairs of real blurred images and corresponding ground truth images obtained by high-speed cameras.
NH-HAZE
Download address: http://m6z.cn/5tyN0D
This is a non-uniform real dataset with pairs of real haze and corresponding haze-free images. This is the first non-homogeneous image deblurring dataset, containing 55 outdoor scenes. Non-uniform fog was introduced into the scene, using a professional fog generator to simulate the real conditions of the fog scene.
ExDark image dataset
Download address: http://suo.nz/2lidoI
The Exclusively Dark (ExDARK) dataset is a collection of 7,363 low-light images from extreme low-light environments to dusk (i.e. 10 different conditions), with 12 object classes (similar to PASCAL VOC), at the image class level and local objects Annotated boxes are added to the borders.
WoodScape autonomous driving fisheye dataset
Download address: http://suo.nz/2HMEtL
WoodScape contains four surround-view cameras and nine tasks, including segmentation, depth estimation, 3D bounding box detection, and novel contamination detection. Provides instance-level semantic annotations of 40 classes for over 10,000 images.
PolyU dataset
Data set download address: https://sourl.cn/rMsdE8
Most previous image denoising methods focus on additive white Gaussian noise (AWGN). However, with the advancement of computer vision technology, the problem of noisy image denoising in the real world has also arisen. To facilitate research on this problem while implementing concurrent real-world image denoising datasets, the authors constructed a new benchmark dataset containing synthetic real-world noisy images of different natural scenes. The images were taken by different cameras with different camera settings.