24 datasets found

The OHaze dataset used for testing the PF-DCP algorithm.

The dataset used for training and testing the pyramid fusion dark channel prior (PF-DCP) algorithm for single image dehazing.

The WeatherStream dataset is a dataset for single image deweathering.

The dataset for non-homogeneous dehazing, which includes RESIDE (Indoor Training Set, ITS), NH-HAZE, NH-HAZE2, and HD-NH-HAZE.

The dataset for non-homogeneous dehazing, which includes RESIDE (Indoor Training Set, ITS), NH-HAZE, NH-HAZE2, and HD-NH-HAZE.

Real-world image dehazing via unsupervised contrastive learning

RealNightHaze is a real-world night dehazing dataset. It includes 440 night hazy images, where 150 images are from [98], 200 images are from [87] and the rest images are...

Nighttime haze images often suffer from reduced visibility. In addition to the common issues faced by night images, such as low light, noise, uneven light distribution, and...

Contrastive learning for compact single image dehazing

From synthetic to real: Image dehazing collaborating with unlabeled real data

Single image dehazing is a challenging ill-posed problem which estimates latent haze-free images from observed hazy images.

The dataset used for testing the proposed T-Net and Stack T-Net for single image dehazing.

The dataset used for training and testing the proposed T-Net and Stack T-Net for single image dehazing.

The Hazy-COCO dataset is a synthetic dataset generated by adding haze to the MS COCO dataset.

The RESIDE dataset is a large-scale dataset for benchmarking single image dehazing algorithms, and it includes both indoor and outdoor hazy images.

The NTIRE2018-Dehazing challenge dataset contains indoor and outdoor hazy images.

The HazeRD dataset contains 75 synthesized hazy images with realistic haze conditions.

The RESIDE dataset contains both synthesized and real-world hazy/clean image pairs of indoor and outdoor scenes.

The proposed Multi-Scale Boosted Dehazing Network with Dense Feature Fusion for image dehazing.

The authors created a custom dataset for their experiment, consisting of 33,000 images of 320 possible object-image combinations, with 10 possible shapes, 8 possible colors, 2...