Existing dehazing methods face challenges in generalization due to the lack of paired real-world training data and tailored models. Recently, some semi-supervised/unsupervised schemes have been explored, achieving impressive performance. However, their performance still depends heavily on synthetic training data and the introduced prior-based strong constraints do not always hold. In this paper, we first introduce RealHQ-HAZE, a new dataset with 200 collected real-world hazy images, 200 corresponding carefully rendered haze-free images, and an additional 1000 varicolored hazy images transferred from the collected images. We also propose a prior-compensated multi-stage dehazing network, PMDN, which can learn different levels of real-world haze distribution through multi-stage progressive learning. To utilize prior knowledge effectively, we introduce a prior-based feature compensation module, guiding intermediate results with an adaptive weight. Additionally, we propose a MixCut consistent dehazing strategy to mix paired and derived images using a cross-cutting scheme, reinforcing dehazing through consistency principles. Extensive experiments demonstrate the effectiveness of our dataset and the superiority of PMDN compared to existing state-of-the-art dehazing methods.

Recent years have witnessed a few attempts of vision transformers for single image super-resolution (SISR). Since the high resolution of intermediate features in SISR models increases memory and computational requirements, efficient SISR transformers are more favored. Based on some popular transformer backbone, many methods have explored reasonable schemes to reduce the computational complexity of the self-attention module while achieving impressive performance. However, these methods only focus on the performance on the training platform (e.g., Pytorch/Tensorflow) without further optimization for the deployment platform (e.g., TensorRT). Therefore, they inevitably contain some redundant operators, posing challenges for subsequent deployment in real-world applications. In this paper, we propose a deployment-friendly transformer unit, namely UFONE (i.e., UnFolding ONce is Enough), to alleviate these problems. In each UFONE, we introduce an Inner-patch Transformer Layer (ITL) to efficiently reconstruct the local structural information from patches and a Spatial-Aware Layer (SAL) to exploit the long-range dependencies between patches. Based on UFONE, we propose a Deployment-friendly Inner-patch Transformer Network (DITN) for the SISR task, which can achieve favorable performance with low latency and memory usage on both training and deployment platforms. Furthermore, to further boost the deployment efficiency of the proposed DITN on TensorRT, we also provide an efficient substitution for layer normalization and propose a fusion optimization strategy for specific operators. Extensive experiments show that our models can achieve competitive results in terms of qualitative and quantitative performance with high deployment efficiency.