Super-resolution (SR), a classical inverse problem in computer vision, is inherently ill-posed, inducing a distribution of plausible solutions for every input. However, the desired result is not simply the expectation of this distribution, which is the blurry image obtained by minimizing pixelwise error, but rather the sample with the highest image quality. A variety of techniques, from perceptual metrics to adversarial losses, are employed to this end. In this work, we explore an alternative: utilizing powerful non-reference image quality assessment (NR-IQA) models in the SR context. We begin with a comprehensive analysis of NR-IQA metrics on human-derived SR data, identifying both the accuracy (human alignment) and complementarity of different metrics. Then, we explore two methods of applying NR-IQA models to SR learning: (i) altering data sampling, by building on an existing multi-ground-truth SR framework, and (ii) directly optimizing a differentiable quality score. Our results demonstrate a more human-centric perception-distortion tradeoff, focusing less on non-perceptual pixelwise distortion, instead improving the balance between perceptual fidelity and human-tuned NR-IQA measures.

Blind face restoration methods have shown remarkable performance, particularly when trained on large-scale synthetic datasets with supervised learning. These datasets are often generated by simulating low-quality face images with a handcrafted image degradation pipeline. The models trained on such synthetic degradations, however, can-not deal with inputs of unseen degradations. In this paper, we address this issue by using only a set of input images, with unknown degradations and without ground truth targets, to fine-tune a restoration model that learns to map them to clean and contextually consistent outputs. We utilize a pre-trained diffusion model as a generative prior through which we generate high quality images from the natural image distribution while maintaining the input image content through consistency constraints. These generated images are then used as pseudo targets to fine-tune a pre-trained restoration model. Unlike many recent approaches that employ diffusion models at test time, we only do so during training and thus maintain an efficient inference-time performance. Extensive experiments show that the proposed approach can consistently improve the perceptual quality of pre-trained blind face restoration models while maintaining great consistency with the input contents. Our best model also achieves the state-of-the-art results on both synthetic and real-world datasets. Project Page.

Neural Radiance Fields (NeRFs) have emerged as a popular approach for novel view synthesis. While NeRFs are quickly being adapted for a wider set of applications, intuitively editing NeRF scenes is still an open challenge. One important editing task is the removal of unwanted objects from a 3D scene, such that the replaced region is visually plausible and consistent with its context. We refer to this task as 3D inpainting. In 3D, solutions must be both consistent across multiple views and geometrically valid. In this paper, we propose a novel 3D inpainting method that addresses these challenges. Given a small set of posed images and sparse annotations in a single input image, our framework first rapidly obtains a 3D segmentation mask for a target object. Using the mask, a perceptual optimization-based approach is then introduced that leverages learned 2D image inpainters, distilling their information into 3D space, while ensuring view consistency. We also address the lack of a diverse benchmark for evaluating 3D scene inpainting methods by introducing a dataset comprised of challenging real-world scenes. In particular, our dataset contains views of the same scene with and without a target object, enabling more principled benchmarking of the 3D inpainting task. We first demonstrate the superiority of our approach on multiview segmentation, comparing to NeRF-based methods and 2D segmentation approaches. We then evaluate on the task of 3D inpainting, establishing state-of-the-art performance against other NeRF manipulation algorithms, as well as a strong 2D image inpainter baseline.

The popularity of Neural Radiance Fields (NeRFs) for view synthesis has led to a desire for NeRF editing tools. Here, we focus on inpainting regions in a view-consistent and controllable manner. In addition to the typical NeRF inputs and masks delineating the unwanted region in each view, we require only a single inpainted view of the scene, i.e., a reference view. We use monocular depth estimators to back-project the inpainted view to the correct 3D positions. Then, via a novel rendering technique, a bilateral solver can construct view-dependent effects in non-reference views, making the inpainted region appear consistent from any view. For non-reference disoccluded regions, which cannot be supervised by the single reference view, we devise a method based on image inpainters to guide both the geometry and appearance. Our approach shows superior performance to NeRF inpainting baselines, with the additional advantage that a user can control the output via a single inpainted image. Please visit our project page.

Taking photographs "in-the-wild" is often hindered by fence obstructions that stand between the camera user and the scene of interest, and which are hard or impossible to avoid. De-fencing is the algorithmic process of automatically removing such obstructions from images, revealing the invisible parts of the scene. While this problem can be formulated as a combination of fence segmentation and image inpainting, this often leads to implausible hallucinations of the occluded regions. Existing multi-frame approaches rely on propagating information to a selected keyframe from its temporal neighbors, but they are often inefficient and struggle with alignment of severely obstructed images. In this work we draw inspiration from the video completion literature, and develop a simplified framework for multi-frame de-fencing that computes high quality flow maps directly from obstructed frames, and uses them to accurately align frames. Our primary focus is efficiency and practicality in a real world setting: the input to our algorithm is a short image burst (5 frames) – a data modality commonly available in modern smartphones– and the output is a single reconstructed keyframe, with the fence removed. Our approach leverages simple yet effective CNN modules, trained on carefully generated synthetic data, and outperforms more complicated alternatives real bursts, both quantitatively and qualitatively, while running real-time.

Many flagship smartphone cameras now use a dedicated neural image signal processor (ISP) to render noisy raw sensor images to the final processed output. Training night-mode ISP networks relies on large-scale datasets of image pairs with: (1) a noisy raw image captured with a short exposure and a high ISO gain; and (2) a ground truth low-noise raw image captured with a long exposure and low ISO that has been rendered through the ISP. Capturing such image pairs is tedious and time-consuming, requiring careful setup to ensure alignment between the image pairs. In addition, ground truth images are often prone to motion blur due to the long exposure. To address this problem, we propose a method that synthesizes nighttime images from day-time images. Daytime images are easy to capture, exhibit low-noise (even on smartphone cameras) and rarely suffer from motion blur. We outline a processing framework to convert daytime raw images to have the appearance of realistic nighttime raw images with different levels of noise. Our procedure allows us to easily produce aligned noisy and clean nighttime image pairs. We show the effectiveness of our synthesis framework by training neural ISPs for nightmode rendering. Furthermore, we demonstrate that using our synthetic nighttime images together with small amounts of real data (e.g., 5% to 10%) yields performance almost on par with training exclusively on real nighttime images. Our dataset and code are available at https://github.com/SamsungLabs/day-to-night.

Modern generative adversarial networks (GANs) predominantly use piecewise linear activation functions in discriminators (or critics), including ReLU and LeakyReLU. Such models learn piecewise linear mappings, where each piece handles a subset of the input space, and the gradients per subset are piecewise constant. Under such a class of discriminator (or critic) functions, we present Gradient Normalization (GraN), a novel input-dependent normalization method, which guarantees a piecewise K-Lipschitz constraint in the input space. In contrast to spectral normalization, GraN does not constrain processing at the individual network layers, and, unlike gradient penalties, strictly enforces a piecewise Lipschitz constraint almost everywhere. Empirically, we demonstrate improved image generation performance across multiple datasets (incl. CIFAR-10/100, STL-10, LSUN bedrooms, and CelebA), GAN loss functions, and metrics. Further, we analyze altering the often untuned Lipschitz constant in several standard GANs, not only attaining significant performance gains, but also finding connections between and training dynamics, particularly in low-gradient loss plateaus, with the common Adam optimizer.

For humans, visual understanding is inherently generative: given a 3D shape, we can postulate how it would look in the world; given a 2D image, we can infer the 3D structure that likely gave rise to it. We can thus translate between the 2D visual and 3D structural modalities of a given object. In the context of computer vision, this corresponds to a learnable module that serves two purposes: (i) generate a realistic rendering of a 3D object (shape-to-image translation) and (ii) infer a realistic 3D shape from an image (image-to-shape translation). In this paper, we learn such a module while being conscious of the difficulties in obtaining large paired 2D-3D datasets. By leveraging generative domain translation methods, we are able to define a learning algorithm that requires only weak supervision, with unpaired data. The resulting model is not only able to perform 3D shape, pose, and texture inference from 2D images, but can also generate novel textured 3D shapes and renders, similar to a graphics pipeline. More specifically, our method (i) infers an explicit 3D mesh representation, (ii) utilizes example shapes to regularize inference, (iii) requires only an image mask (no keypoints or camera extrinsics), and (iv) has generative capabilities. While prior work explores subsets of these properties, their combination is novel. We demonstrate the utility of our learned representation, as well as its performance on image generation and unpaired 3D shape inference tasks.

Beginning with R-CNN, there has been a rapid advancement in two-stage object detection approaches. While two-stage approaches remain the state-of-the-art in object detection, anchor-free single-stage methods have been gaining momentum. We believe that the strength of the former is in their region of interest (ROI) pooling stage, while the latter simplifies the learning problem by converting object detection into dense per-pixel prediction tasks. In this paper, we propose to combine the strengths of each approach in a new architecture. In particular, we first define several auxiliary tasks related to object detection and generate dense per-pixel predictions using a shared feature extraction backbone. As a consequence of this architecture, the shared backbone is trained using both the standard object detection losses and these per-pixel ones. Moreover, by combining the features from dense predictions with those from the backbone, we realize a more discriminative representation for subsequent downstream processing. In addition, we feed the fused features into a novel multi-scale ROI pooling layer, followed by per-ROI predictions. We refer to our architecture as the Dense Auxiliary Tasks Network (DATNet). We present an extensive set of evaluations of our method on the Pascal VOC and COCO datasets and show considerable accuracy improvements over comparable baselines.

Skeletonization algorithms typically decompose an object's silhouette into a set of symmetric parts, offering a powerful representation for shape categorization. However, having access to an object's silhouette assumes correct figure-ground segmentation, leading to a disconnect with the mainstream categorization community, which attempts to recognize objects from cluttered images. In this paper, we present a novel approach to recovering and grouping the symmetric parts of an object from a cluttered scene. We begin by using a multiresolution superpixel segmentation to generate medial point hypotheses, and use a learned affinity function to perceptually group nearby medial points likely to belong to the same medial branch. In the next stage, we learn higher granularity affinity functions to group the resulting medial branches likely to belong to the same object. The resulting framework yields a skeletal approximation that's free of many of the instabilities plaguing traditional skeletons. More importantly, it doesn't require a closed contour, enabling the application of skeleton-based categorization systems to more realistic imagery

We describe a geometric-flow-based algorithm for computing a dense oversegmentation of an image, often referred to as superpixels. It produces segments that, on one hand, respect local image boundaries, while, on the other hand, limiting undersegmentation through a compactness constraint. It is very fast, with complexity that is approximately linear in image size, and can be applied to megapixel sized images with high superpixel densities in a matter of minutes. We show qualitative demonstrations of high-quality results on several complex images. The Berkeley database is used to quantitatively compare its performance to a number of oversegmentation algorithms, showing that it yields less undersegmentation than algorithms that lack a compactness constraint while offering a significant speedup over N-cuts, which does enforce compactness.