Although gaze estimation methods have been developed with deep learning techniques, there has been no such approach as aim to attain accurate performance in low-resolution face images with a pixel width of 50 pixels or less. To solve a limitation under the challenging low-resolution conditions, we propose a high-frequency attentive super-resolved gaze estimation network, i.e., HAZE-Net. Our network improves the resolution of the input image and enhances the eye features and those boundaries via a proposed super-resolution module based on a high-frequency attention block. In addition, our gaze estimation module utilizes high-frequency components of the eye as well as the global appearance map. We also utilize the structural location information of faces to approximate head pose. The experimental results indicate that the proposed method exhibits robust gaze estimation performance even in low-resolution face images with 28x28 pixels. The source code of this work is available at https://github.com/dbseorms16/HAZE_Net/.
Video conferencing systems suffer from poor user experience when network conditions deteriorate because current video codecs simply cannot operate at extremely low bitrates. Recently, several neural alternatives have been proposed that reconstruct talking head videos at very low bitrates using sparse representations of each frame such as facial landmark information. However, these approaches produce poor reconstructions in scenarios with major movement or occlusions over the course of a call, and do not scale to higher resolutions. We design Gemino, a new neural compression system for video conferencing based on a novel high-frequency-conditional super-resolution pipeline. Gemino upsamples a very low-resolution version of each target frame while enhancing high-frequency details (e.g., skin texture, hair, etc.) based on information extracted from a single high-resolution reference image. We use a multi-scale architecture that runs different components of the model at different resolutions, allowing it to scale to resolutions comparable to 720p, and we personalize the model to learn specific details of each person, achieving much better fidelity at low bitrates. We implement Gemino atop aiortc, an open-source Python implementation of WebRTC, and show that it operates on 1024x1024 videos in real-time on a Titan X GPU, and achieves 2.2-5x lower bitrate than traditional video codecs for the same perceptual quality.
The process of obtaining high-resolution images from single or multiple low-resolution images of the same scene is of great interest for real-world image and signal processing applications. This study is about exploring the potential usage of deep learning based image super-resolution algorithms on thermal data for producing high quality thermal imaging results for in-cabin vehicular driver monitoring systems. In this work we have proposed and developed a novel multi-image super-resolution recurrent neural network to enhance the resolution and improve the quality of low-resolution thermal imaging data captured from uncooled thermal cameras. The end-to-end fully convolutional neural network is trained from scratch on newly acquired thermal data of 30 different subjects in indoor environmental conditions. The effectiveness of the thermally tuned super-resolution network is validated quantitatively as well as qualitatively on test data of 6 distinct subjects. The network was able to achieve a mean peak signal to noise ratio of 39.24 on the validation dataset for 4x super-resolution, outperforming bicubic interpolation both quantitatively and qualitatively.
Fetal brain magnetic resonance imaging serves as an emerging modality for prenatal counseling and diagnosis in disorders affecting the brain. Machine learning based segmentation plays an important role in the quantification of brain development. However, a limiting factor is the lack of sufficiently large, labeled training data. Our study explored the application of SPADE, a conditional general adversarial network (cGAN), which learns the mapping from the label to the image space. The input to the network was super-resolution T2-weighted cerebral MRI data of 120 fetuses (gestational age range: 20-35 weeks, normal and pathological), which were annotated for 7 different tissue categories. SPADE networks were trained on 256*256 2D slices of the reconstructed volumes (image and label pairs) in each orthogonal orientation. To combine the generated volumes from each orientation into one image, a simple mean of the outputs of the three networks was taken. Based on the label maps only, we synthesized highly realistic images. However, some finer details, like small vessels were not synthesized. A structural similarity index (SSIM) of 0.972+-0.016 and correlation coefficient of 0.974+-0.008 were achieved. To demonstrate the capacity of the cGAN to create new anatomical variants, we artificially dilated the ventricles in the segmentation map and created synthetic MRI of different degrees of fetal hydrocephalus. cGANs, such as the SPADE algorithm, allow the generation of hypothetically unseen scenarios and anatomical configurations in the label space, which data in turn can be utilized for training various machine learning algorithms. In the future, this algorithm would be used for generating large, synthetic datasets representing fetal brain development. These datasets would potentially improve the performance of currently available segmentation networks.
High-quality HDRIs(High Dynamic Range Images), typically HDR panoramas, are one of the most popular ways to create photorealistic lighting and 360-degree reflections of 3D scenes in graphics. Given the difficulty of capturing HDRIs, a versatile and controllable generative model is highly desired, where layman users can intuitively control the generation process. However, existing state-of-the-art methods still struggle to synthesize high-quality panoramas for complex scenes. In this work, we propose a zero-shot text-driven framework, Text2Light, to generate 4K+ resolution HDRIs without paired training data. Given a free-form text as the description of the scene, we synthesize the corresponding HDRI with two dedicated steps: 1) text-driven panorama generation in low dynamic range(LDR) and low resolution, and 2) super-resolution inverse tone mapping to scale up the LDR panorama both in resolution and dynamic range. Specifically, to achieve zero-shot text-driven panorama generation, we first build dual codebooks as the discrete representation for diverse environmental textures. Then, driven by the pre-trained CLIP model, a text-conditioned global sampler learns to sample holistic semantics from the global codebook according to the input text. Furthermore, a structure-aware local sampler learns to synthesize LDR panoramas patch-by-patch, guided by holistic semantics. To achieve super-resolution inverse tone mapping, we derive a continuous representation of 360-degree imaging from the LDR panorama as a set of structured latent codes anchored to the sphere. This continuous representation enables a versatile module to upscale the resolution and dynamic range simultaneously. Extensive experiments demonstrate the superior capability of Text2Light in generating high-quality HDR panoramas. In addition, we show the feasibility of our work in realistic rendering and immersive VR.
Diabetic foot ulcers (DFUs) constitute a serious complication for people with diabetes. The care of DFU patients can be substantially improved through self-management, in order to achieve early-diagnosis, ulcer prevention, and complications management in existing ulcers. In this paper, we investigate two categories of image-to-image translation techniques (ItITT), which will support decision making and monitoring of diabetic foot ulcers: noise reduction and super-resolution. In the former case, we investigated the capabilities on noise removal, for convolutional neural network stacked-autoencoders (CNN-SAE). CNN-SAE was tested on RGB images, induced with Gaussian noise. The latter scenario involves the deployment of four deep learning super-resolution models. The performance of all models, for both scenarios, was evaluated in terms of execution time and perceived quality. Results indicate that applied techniques consist a viable and easy to implement alternative that should be used by any system designed for DFU monitoring.
Deep-learning-based hyperspectral image (HSI) restoration methods have gained great popularity for their remarkable performance but often demand expensive network retraining whenever the specifics of task changes. In this paper, we propose to restore HSIs in a unified approach with an effective plug-and-play method, which can jointly retain the flexibility of optimization-based methods and utilize the powerful representation capability of deep neural networks. Specifically, we first develop a new deep HSI denoiser leveraging gated recurrent convolution units, short- and long-term skip connections, and an augmented noise level map to better exploit the abundant spatio-spectral information within HSIs. It, therefore, leads to the state-of-the-art performance on HSI denoising under both Gaussian and complex noise settings. Then, the proposed denoiser is inserted into the plug-and-play framework as a powerful implicit HSI prior to tackle various HSI restoration tasks. Through extensive experiments on HSI super-resolution, compressed sensing, and inpainting, we demonstrate that our approach often achieves superior performance, which is competitive with or even better than the state-of-the-art on each task, via a single model without any task-specific training.
Assuming a known degradation model, the performance of a learned image super-resolution (SR) model depends on how well the variety of image characteristics within the training set matches those in the test set. As a result, the performance of an SR model varies noticeably from image to image over a test set depending on whether characteristics of specific images are similar to those in the training set or not. Hence, in general, a single SR model cannot generalize well enough for all types of image content. In this work, we show that training multiple SR models for different classes of images (e.g., for text, texture, etc.) to exploit class-specific image priors and employing a post-processing network that learns how to best fuse the outputs produced by these multiple SR models surpasses the performance of state-of-the-art generic SR models. Experimental results clearly demonstrate that the proposed multiple-model SR (MMSR) approach significantly outperforms a single pre-trained state-of-the-art SR model both quantitatively and visually. It even exceeds the performance of the best single class-specific SR model trained on similar text or texture images.
Flow-based generative super-resolution (SR) models learn to produce a diverse set of feasible SR solutions, called the SR space. Diversity of SR solutions increases with the temperature ($\tau$) of latent variables, which introduces random variations of texture among sample solutions, resulting in visual artifacts and low fidelity. In this paper, we present a simple but effective image ensembling/fusion approach to obtain a single SR image eliminating random artifacts and improving fidelity without significantly compromising perceptual quality. We achieve this by benefiting from a diverse set of feasible photo-realistic solutions in the SR space spanned by flow models. We propose different image ensembling and fusion strategies which offer multiple paths to move sample solutions in the SR space to more desired destinations in the perception-distortion plane in a controllable manner depending on the fidelity vs. perceptual quality requirements of the task at hand. Experimental results demonstrate that our image ensembling/fusion strategy achieves more promising perception-distortion trade-off compared to sample SR images produced by flow models and adversarially trained models in terms of both quantitative metrics and visual quality.
Benefiting from the vigorous development of deep learning, many CNN-based image super-resolution methods have emerged and achieved better results than traditional algorithms. However, it is difficult for most algorithms to adaptively adjust the spatial region and channel features at the same time, let alone the information exchange between them. In addition, the exchange of information between attention modules is even less visible to researchers. To solve these problems, we put forward a lightweight spatial-channel adaptive coordination of multilevel refinement enhancement networks(MREN). Specifically, we construct a space-channel adaptive coordination block, which enables the network to learn the spatial region and channel feature information of interest under different receptive fields. In addition, the information of the corresponding feature processing level between the spatial part and the channel part is exchanged with the help of jump connection to achieve the coordination between the two. We establish a communication bridge between attention modules through a simple linear combination operation, so as to more accurately and continuously guide the network to pay attention to the information of interest. Extensive experiments on several standard test sets have shown that our MREN achieves superior performance over other advanced algorithms with a very small number of parameters and very low computational complexity.
This paper studies the estimation of cascaded channels in passive intelligent reflective surface (IRS)- aided multiple-input multiple-output (MIMO) systems employing hybrid precoders and combiners. We propose a low-complexity solution that estimates the channel parameters progressively. The angles of departure (AoDs) and angles of arrival (AoAs) at the transmitter and receiver, respectively, are first estimated using inductive matrix completion (IMC) followed by root-MUSIC based super-resolution spectrum estimation. Forward-backward spatial smoothing (FBSS) is applied to address the coherence issue. Using the estimated AoAs and AoDs, the training precoders and combiners are then optimized and the angle differences between the AoAs and AoDs at the IRS are estimated using the least squares (LS) method followed by FBSS and the root-MUSIC algorithm. Finally, the composite path gains of the cascaded channel are estimated using on-grid sparse recovery with a small-size dictionary. The simulation results suggest that the proposed estimator can achieve improved channel parameter estimation performance with lower complexity as compared to several recently reported alternatives, thanks to the exploitation of the knowledge of the array responses and low-rankness of the channel using low-complexity algorithms at all the stages.
Inferring the timing and amplitude of perturbations in epidemiological systems from their stochastically spread low-resolution outcomes is as relevant as challenging. It is a requirement for current approaches to overcome the need to know the details of the perturbations to proceed with the analyses. However, the general problem of connecting epidemiological curves with the underlying incidence lacks the highly effective methodology present in other inverse problems, such as super-resolution and dehazing from computer vision. Here, we develop an unsupervised physics-informed convolutional neural network approach in reverse to connect death records with incidence that allows the identification of regime changes at single-day resolution. Applied to COVID-19 data with proper regularization and model-selection criteria, the approach can identify the implementation and removal of lockdowns and other nonpharmaceutical interventions with 0.93-day accuracy over the time span of a year.
Formaldehyde (HCHO) is one one of the most important trace gas in the atmosphere, as it is a pollutant causing respiratory and other diseases. It is also a precursor of tropospheric ozone which damages crops and deteriorates human health. Study of HCHO chemistry and long-term monitoring using satellite data is important from the perspective of human health, food security and air pollution. Dynamic atmospheric chemistry models struggle to simulate atmospheric formaldehyde and often overestimate by up to two times relative to satellite observations and reanalysis. Spatial distribution of modelled HCHO also fail to match satellite observations. Here, we present deep learning approach using a simple super-resolution based convolutional neural network towards simulating fast and reliable atmospheric HCHO. Our approach is an indirect method of HCHO estimation without the need to chemical equations. We find that deep learning outperforms dynamical model simulations which involves complicated atmospheric chemistry representation. Causality establishing the nonlinear relationships of different variables to target formaldehyde is established in our approach by using a variety of precursors from meteorology and chemical reanalysis to target OMI AURA satellite based HCHO predictions. We choose South Asia for testing our implementation as it doesnt have in situ measurements of formaldehyde and there is a need for improved quality data over the region. Moreover, there are spatial and temporal data gaps in the satellite product which can be removed by trustworthy modelling of atmospheric formaldehyde. This study is a novel attempt using computer vision for trustworthy modelling of formaldehyde from remote sensing can lead to cascading societal benefits.
Urban downscaling is a link to transfer the knowledge from coarser climate information to city scale assessments. These high-resolution assessments need multiyear climatology of past data and future projections, which are complex and computationally expensive to generate using traditional numerical weather prediction models. The city of Austin, Texas, USA has seen tremendous growth in the past decade. Systematic planning for the future requires the availability of fine resolution city-scale datasets. In this study, we demonstrate a novel approach generating a general purpose operator using deep learning to perform urban downscaling. The algorithm employs an iterative super-resolution convolutional neural network (Iterative SRCNN) over the city of Austin, Texas, USA. We show the development of a high-resolution gridded precipitation product (300 m) from a coarse (10 km) satellite-based product (JAXA GsMAP). High resolution gridded datasets of precipitation offer insights into the spatial distribution of heavy to low precipitation events in the past. The algorithm shows improvement in the mean peak-signal-to-noise-ratio and mutual information to generate high resolution gridded product of size 300 m X 300 m relative to the cubic interpolation baseline. Our results have implications for developing high-resolution gridded-precipitation urban datasets and the future planning of smart cities for other cities and other climatic variables.
Optical coherence tomography (OCT) is a micrometer-scale, volumetric imaging modality that has become a clinical standard in ophthalmology. OCT instruments image by raster-scanning a focused light spot across the retina, acquiring sequential cross-sectional images to generate volumetric data. Patient eye motion during the acquisition poses unique challenges: Non-rigid, discontinuous distortions can occur, leading to gaps in data and distorted topographic measurements. We present a new distortion model and a corresponding fully-automatic, reference-free optimization strategy for computational motion correction in orthogonally raster-scanned, retinal OCT volumes. Using a novel, domain-specific spatiotemporal parametrization of forward-warping displacements, eye motion can be corrected continuously for the first time. Parameter estimation with temporal regularization improves robustness and accuracy over previous spatial approaches. We correct each A-scan individually in 3D in a single mapping, including repeated acquisitions used in OCT angiography protocols. Specialized 3D forward image warping reduces median runtime to < 9 s, fast enough for clinical use. We present a quantitative evaluation on 18 subjects with ocular pathology and demonstrate accurate correction during microsaccades. Transverse correction is limited only by ocular tremor, whereas submicron repeatability is achieved axially (0.51 um median of medians), representing a dramatic improvement over previous work. This allows assessing longitudinal changes in focal retinal pathologies as a marker of disease progression or treatment response, and promises to enable multiple new capabilities such as supersampled/super-resolution volume reconstruction and analysis of pathological eye motion occuring in neurological diseases.
Magnetic resonance imaging (MRI) with high resolution (HR) provides more detailed information for accurate diagnosis and quantitative image analysis. Despite the significant advances, most existing super-resolution (SR) reconstruction network for medical images has two flaws: 1) All of them are designed in a black-box principle, thus lacking sufficient interpretability and further limiting their practical applications. Interpretable neural network models are of significant interest since they enhance the trustworthiness required in clinical practice when dealing with medical images. 2) most existing SR reconstruction approaches only use a single contrast or use a simple multi-contrast fusion mechanism, neglecting the complex relationships between different contrasts that are critical for SR improvement. To deal with these issues, in this paper, a novel Model-Guided interpretable Deep Unfolding Network (MGDUN) for medical image SR reconstruction is proposed. The Model-Guided image SR reconstruction approach solves manually designed objective functions to reconstruct HR MRI. We show how to unfold an iterative MGDUN algorithm into a novel model-guided deep unfolding network by taking the MRI observation matrix and explicit multi-contrast relationship matrix into account during the end-to-end optimization. Extensive experiments on the multi-contrast IXI dataset and BraTs 2019 dataset demonstrate the superiority of our proposed model.
Image retargeting aims at altering an image size while preserving important content and minimizing noticeable distortions. However, previous image retargeting methods create outputs that suffer from artifacts and distortions. Besides, most previous works attempt to retarget the background and foreground of the input image simultaneously. Simultaneous resizing of the foreground and background causes changes in the aspect ratios of the objects. The change in the aspect ratio is specifically not desirable for human objects. We propose a retargeting method that overcomes these problems. The proposed approach consists of the following steps. Firstly, an inpainting method uses the input image and the binary mask of foreground objects to produce a background image without any foreground objects. Secondly, the seam carving method resizes the background image to the target size. Then, a super-resolution method increases the input image quality, and we then extract the foreground objects. Finally, the retargeted background and the extracted super-resolued objects are fed into a particle swarm optimization algorithm (PSO). The PSO algorithm uses aesthetic quality assessment as its objective function to identify the best location and size for the objects to be placed in the background. We used image quality assessment and aesthetic quality assessment measures to show our superior results compared to popular image retargeting techniques.
Channel modeling has always been the core part in communication system design and development, especially in 5G and 6G era. Traditional approaches like stochastic channel modeling and ray-tracing (RT) based channel modeling depend heavily on measurement data or simulation, which are usually expensive and time consuming. In this paper, we propose a novel super resolution (SR) model for generating channel characteristics data. The model is based on multi-task learning (MTL) convolutional neural networks (CNN) with residual connection. Experiments demonstrate that the proposed SR model could achieve excellent performances in mean absolute error and standard deviation of error. Advantages of the proposed model are demonstrated in comparisons with other state-of-the-art deep learning models. Ablation study also proved the necessity of multi-task learning and techniques in model design. The contribution in this paper could be helpful in channel modeling, network optimization, positioning and other wireless channel characteristics related work by largely reducing workload of simulation or measurement.
Single-Image Super-Resolution can support robotic tasks in environments where a reliable visual stream is required to monitor the mission, handle teleoperation or study relevant visual details. In this work, we propose an efficient Generative Adversarial Network model for real-time Super-Resolution, called EdgeSRGAN (code available at https://github.com/PIC4SeR/EdgeSRGAN). We adopt a tailored architecture of the original SRGAN and model quantization to boost the execution on CPU and Edge TPU devices, achieving up to 200 fps inference. We further optimize our model by distilling its knowledge to a smaller version of the network and obtain remarkable improvements compared to the standard training approach. Our experiments show that our fast and lightweight model preserves considerably satisfying image quality compared to heavier state-of-the-art models. Finally, we conduct experiments on image transmission with bandwidth degradation to highlight the advantages of the proposed system for mobile robotic applications.
Magnetic Resonance Imaging (MRI) is important in clinic to produce high resolution images for diagnosis, but its acquisition time is long for high resolution images. Deep learning based MRI super resolution methods can reduce scan time without complicated sequence programming, but may create additional artifacts due to the discrepancy between training data and testing data. Data consistency layer can improve the deep learning results but needs raw k-space data. In this work, we propose a magnitude-image based data consistency deep learning MRI super resolution method to improve super resolution images' quality without raw k-space data. Our experiments show that the proposed method can improve NRMSE and SSIM of super resolution images compared to the same Convolutional Neural Network (CNN) block without data consistency module.