Image colorization is inherently an ill-posed problem with multi-modal uncertainty. Previous methods leverage the deep neural network to map input grayscale images to plausible color outputs directly. Although these learning-based methods have shown impressive performance, they usually fail on the input images that contain multiple objects. The leading cause is that existing models perform learning and colorization on the entire image. In the absence of a clear figure-ground separation, these models cannot effectively locate and learn meaningful object-level semantics. In this paper, we propose a method for achieving instance-aware colorization. Our network architecture leverages an off-the-shelf object detector to obtain cropped object images and uses an instance colorization network to extract object-level features. We use a similar network to extract the full-image features and apply a fusion module to full object-level and image-level features to predict the final colors. Both colorization networks and fusion modules are learned from a large-scale dataset. Experimental results show that our work outperforms existing methods on different quality metrics and achieves state-of-the-art performance on image colorization.
The self-attention mechanism has attracted wide publicity for its most important advantage of modeling long dependency, and its variations in computer vision tasks, the non-local block tries to model the global dependency of the input feature maps. Gathering global contextual information will inevitably need a tremendous amount of memory and computing resources, which has been extensively studied in the past several years. However, there is a further problem with the self-attention scheme: is all information gathered from the global scope helpful for the contextual modelling? To our knowledge, few studies have focused on the problem. Aimed at both questions this paper proposes the salient positions-based attention scheme SPANet, which is inspired by some interesting observations on the attention maps and affinity matrices generated in self-attention scheme. We believe these observations are beneficial for better understanding of the self-attention. SPANet uses the salient positions selection algorithm to select only a limited amount of salient points to attend in the attention map computing. This approach will not only spare a lot of memory and computing resources, but also try to distill the positive information from the transformation of the input feature maps. In the implementation, considering the feature maps with channel high dimensions, which are completely different from the general visual image, we take the squared power of the feature maps along the channel dimension as the saliency metric of the positions. In general, different from the non-local block method, SPANet models the contextual information using only the selected positions instead of all, along the channel dimension instead of space dimension. Our source code is available at https://github.com/likyoo/SPANet.
The recently developed DEtection TRansformer (DETR) establishes a new object detection paradigm by eliminating a series of hand-crafted components. However, DETR suffers from extremely slow convergence, which increases the training cost significantly. We observe that the slow convergence is largely attributed to the complication in matching object queries with target features in different feature embedding spaces. This paper presents SAM-DETR, a Semantic-Aligned-Matching DETR that greatly accelerates DETR's convergence without sacrificing its accuracy. SAM-DETR addresses the convergence issue from two perspectives. First, it projects object queries into the same embedding space as encoded image features, where the matching can be accomplished efficiently with aligned semantics. Second, it explicitly searches salient points with the most discriminative features for semantic-aligned matching, which further speeds up the convergence and boosts detection accuracy as well. Being like a plug and play, SAM-DETR complements existing convergence solutions well yet only introduces slight computational overhead. Extensive experiments show that the proposed SAM-DETR achieves superior convergence as well as competitive detection accuracy. The implementation codes are available at https://github.com/ZhangGongjie/SAM-DETR.
We introduce a novel formulation for guided super-resolution. Its core is a differentiable optimisation layer that operates on a learned affinity graph. The learned graph potentials make it possible to leverage rich contextual information from the guide image, while the explicit graph optimisation within the architecture guarantees rigorous fidelity of the high-resolution target to the low-resolution source. With the decision to employ the source as a constraint rather than only as an input to the prediction, our method differs from state-of-the-art deep architectures for guided super-resolution, which produce targets that, when downsampled, will only approximately reproduce the source. This is not only theoretically appealing, but also produces crisper, more natural-looking images. A key property of our method is that, although the graph connectivity is restricted to the pixel lattice, the associated edge potentials are learned with a deep feature extractor and can encode rich context information over large receptive fields. By taking advantage of the sparse graph connectivity, it becomes possible to propagate gradients through the optimisation layer and learn the edge potentials from data. We extensively evaluate our method on several datasets, and consistently outperform recent baselines in terms of quantitative reconstruction errors, while also delivering visually sharper outputs. Moreover, we demonstrate that our method generalises particularly well to new datasets not seen during training.
Few-shot counting aims to count objects of any class in an image given only a few exemplars of the same class. Existing correlation-based few-shot counting approaches suffer from the coarseness and low semantic level of the correlation. To solve these problems, we propose an iterative framework to progressively refine the exemplar-related features based on the correlation between the image and exemplars. Then the density map is predicted from the final refined feature map. The iterative framework includes a Correlation Distillation module and a Feature Refinement module. During the iterations, the exemplar-related features are gradually refined, while the exemplar-unrelated features are suppressed, benefiting few-shot counting where the exemplar-related features are more important. Our approach surpasses all baselines significantly on few-shot counting benchmark FSC-147. Surprisingly, though designed for general class-agnostic counting, our approach still achieves state-of-the-art performance on car counting benchmarks CARPK and PUCPR+, and crowd counting benchmarks UCSD and Mall. We also achieve competitive performance on crowd counting benchmark ShanghaiTech. The code will be released soon.
Infrared and visible image fusion, a hot topic in the field of image processing, aims at obtaining fused images keeping the advantages of source images. This paper proposes a novel auto-encoder (AE) based fusion network. The core idea is that the encoder decomposes an image into background and detail feature maps with low- and high-frequency information, respectively, and that the decoder recovers the original image. To this end, the loss function makes the background/detail feature maps of source images similar/dissimilar. In the test phase, background and detail feature maps are respectively merged via a fusion module, and the fused image is recovered by the decoder. Qualitative and quantitative results illustrate that our method can generate fusion images containing highlighted targets and abundant detail texture information with strong robustness and meanwhile surpass state-of-the-art (SOTA) approaches.
Synthetic aperture radar (SAR) image change detection is a vital yet challenging task in the field of remote sensing image analysis. Most previous works adopt a self-supervised method which uses pseudo-labeled samples to guide subsequent training and testing. However, deep networks commonly require many high-quality samples for parameter optimization. The noise in pseudo-labels inevitably affects the final change detection performance. To solve the problem, we propose a Graph-based Knowledge Supplement Network (GKSNet). To be more specific, we extract discriminative information from the existing labeled dataset as additional knowledge, to suppress the adverse effects of noisy samples to some extent. Afterwards, we design a graph transfer module to distill contextual information attentively from the labeled dataset to the target dataset, which bridges feature correlation between datasets. To validate the proposed method, we conducted extensive experiments on four SAR datasets, which demonstrated the superiority of the proposed GKSNet as compared to several state-of-the-art baselines. Our codes are available at https://github.com/summitgao/SAR_CD_GKSNet.
Optical Coherence Tomography (OCT) is a widely used non-invasive biomedical imaging modality that can rapidly provide volumetric images of samples. Here, we present a deep learning-based image reconstruction framework that can generate swept-source OCT (SS-OCT) images using undersampled spectral data, without any spatial aliasing artifacts. This neural network-based image reconstruction does not require any hardware changes to the optical set-up and can be easily integrated with existing swept-source or spectral domain OCT systems to reduce the amount of raw spectral data to be acquired. To show the efficacy of this framework, we trained and blindly tested a deep neural network using mouse embryo samples imaged by an SS-OCT system. Using 2-fold undersampled spectral data (i.e., 640 spectral points per A-line), the trained neural network can blindly reconstruct 512 A-lines in ~6.73 ms using a desktop computer, removing spatial aliasing artifacts due to spectral undersampling, also presenting a very good match to the images of the same samples, reconstructed using the full spectral OCT data (i.e., 1280 spectral points per A-line). We also successfully demonstrate that this framework can be further extended to process 3x undersampled spectral data per A-line, with some performance degradation in the reconstructed image quality compared to 2x spectral undersampling. This deep learning-enabled image reconstruction approach can be broadly used in various forms of spectral domain OCT systems, helping to increase their imaging speed without sacrificing image resolution and signal-to-noise ratio.
Domain Adaptive Object Detection (DAOD) leverages a labeled domain to learn an object detector generalizing to a novel domain free of annotations. Recent advances align class-conditional distributions by narrowing down cross-domain prototypes (class centers). Though great success,they ignore the significant within-class variance and the domain-mismatched semantics within the training batch, leading to a sub-optimal adaptation. To overcome these challenges, we propose a novel SemantIc-complete Graph MAtching (SIGMA) framework for DAOD, which completes mismatched semantics and reformulates the adaptation with graph matching. Specifically, we design a Graph-embedded Semantic Completion module (GSC) that completes mismatched semantics through generating hallucination graph nodes in missing categories. Then, we establish cross-image graphs to model class-conditional distributions and learn a graph-guided memory bank for better semantic completion in turn. After representing the source and target data as graphs, we reformulate the adaptation as a graph matching problem, i.e., finding well-matched node pairs across graphs to reduce the domain gap, which is solved with a novel Bipartite Graph Matching adaptor (BGM). In a nutshell, we utilize graph nodes to establish semantic-aware node affinity and leverage graph edges as quadratic constraints in a structure-aware matching loss, achieving fine-grained adaptation with a node-to-node graph matching. Extensive experiments verify that SIGMA outperforms existing works significantly. Our code is available at https://github.com/CityU-AIM-Group/SIGMA.
A camera begins to sense light the moment we press the shutter button. During the exposure interval, relative motion between the scene and the camera causes motion blur, a common undesirable visual artifact. This paper presents E-CIR, which converts a blurry image into a sharp video represented as a parametric function from time to intensity. E-CIR leverages events as an auxiliary input. We discuss how to exploit the temporal event structure to construct the parametric bases. We demonstrate how to train a deep learning model to predict the function coefficients. To improve the appearance consistency, we further introduce a refinement module to propagate visual features among consecutive frames. Compared to state-of-the-art event-enhanced deblurring approaches, E-CIR generates smoother and more realistic results. The implementation of E-CIR is available at https://github.com/chensong1995/E-CIR.