Deep Image Harmonization via Domain Verification

Image composition is an important operation in image processing, but the inconsistency between foreground and background significantly degrades the quality of composite image. Image harmonization, aiming to make the foreground compatible with the background, is a promising yet challenging task. However, the lack of high-quality publicly available dataset for image harmonization greatly hinders the development of image harmonization techniques. In this work, we contribute an image harmonization dataset by generating synthesized composite images based on COCO (resp., Adobe5k, Flickr, day2night) dataset, leading to our HCOCO (resp., HAdobe5k, HFlickr, Hday2night) sub-dataset. Moreover, we propose a new deep image harmonization method with a novel domain verification discriminator, enlightened by the following insight. Specifically, incompatible foreground and background belong to two different domains, so we need to translate the domain of foreground to the same domain as background. Our proposed domain verification discriminator can play such a role by pulling close the domains of foreground and background. Extensive experiments on our constructed dataset demonstrate the effectiveness of our proposed method. Our dataset is released in https://github.com/bcmi/Image_Harmonization_Datasets.

Encoding Selection for Solving Hamiltonian Cycle Problems with ASP

It is common for search and optimization problems to have alternative equivalent encodings in ASP. Typically none of them is uniformly better than others when evaluated on broad classes of problem instances. We claim that one can improve the solving ability of ASP by using machine learning techniques to select encodings likely to perform well on a given instance. We substantiate this claim by studying the hamiltonian cycle problem. We propose several equivalent encodings of the problem and several classes of hard instances. We build models to predict the behavior of each encoding, and then show that selecting encodings for a given instance using the learned performance predictors leads to significant performance gains.

Quantum algorithm for finding the negative curvature direction in non-convex optimization

We present an efficient quantum algorithm aiming to find the negative curvature direction for escaping the saddle point, which is the critical subroutine for many second-order non-convex optimization algorithms. We prove that our algorithm could produce the target state corresponding to the negative curvature direction with query complexity O(polylog(d) /{\epsilon}), where d is the dimension of the optimization function. The quantum negative curvature finding algorithm is exponentially faster than any known classical method which takes time at least O(d /\sqrt{\epsilon}). Moreover, we propose an efficient quantum algorithm to achieve the classical read-out of the target state. Our classical read-out algorithm runs exponentially faster on the degree of d than existing counterparts.

NODE: Extreme Low Light Raw Image Denoising using a Noise Decomposition Network

Denoising extreme low light images is a challenging task due to the high noise level. When the illumination is low, digital cameras increase the ISO (electronic gain) to amplify the brightness of captured data. However, this in turn amplifies the noise, arising from read, shot, and defective pixel sources. In the raw domain, read and shot noise are effectively modelled using Gaussian and Poisson distributions respectively, whereas defective pixels can be modeled with impulsive noise. In extreme low light imaging, noise removal becomes a critical challenge to produce a high quality, detailed image with low noise. In this paper, we propose a multi-task deep neural network called Noise Decomposition (NODE) that explicitly and separately estimates defective pixel noise, in conjunction with Gaussian and Poisson noise, to denoise an extreme low light image. Our network is purposely designed to work with raw data, for which the noise is more easily modeled before going through non-linear transformations in the image signal processing (ISP) pipeline. Quantitative and qualitative evaluation show the proposed method to be more effective at denoising real raw images than state-of-the-art techniques.

Image Harmonization Datasets: HCOCO, HAdobe5k, HFlickr, and Hday2night

Image composition is an important operation in image processing, but the inconsistency between foreground and background significantly degrades the quality of composite image. Image harmonization, which aims to make the foreground compatible with the background, is a promising yet challenging task. However, the lack of high-quality public dataset for image harmonization, which significantly hinders the development of image harmonization techniques. Therefore, we create synthesized composite images based on existing COCO (resp., Adobe5k, day2night) dataset, leading to our HCOCO (resp., HAdobe5k, Hday2night) dataset. To enrich the diversity of our datasets, we also generate synthesized composite images based on our collected Flick images, leading to our HFlickr dataset. All four datasets are released in https://github.com/bcmi/Image_Harmonization_Datasets.

Optimal Feature Transport for Cross-View Image Geo-Localization

This paper addresses the problem of cross-view image based localization, where the geographic location of a ground-level street-view query image is estimated by matching it against a large scale aerial map (e.g., a high-resolution satellite image). State-of-the-art deep-learning based methods tackle this problem as deep metric learning which aims to learn global feature representations of the scene seen by the two different views. Despite promising results are obtained by such deep metric learning methods, they, however, fail to exploit a crucial cue relevant for localization, namely, the spatial layout of local features. Moreover, little attention is paid to the obvious domain gap (between aerial view and ground view) in the context of cross-view localization. This paper proposes a novel Cross-View Feature Transport (CVFT) technique to explicitly establish cross-view domain transfer that facilitates feature alignment between ground and aerial images. Specifically, we implement the CVFT as a network layer, which transports features from one domain to the other, leading to more meaningful feature similarity comparison. Our model is differentiable and can be learned end-to-end. Experiments on large-scale datasets have demonstrated that our method has remarkably boosted the state-of-the-art cross-view localization performance, e.g., on the CVUSA dataset, with significant improvements for top-1 recall from 40.79% to 61.43%, and for top-10 from 76.36% to 90.49%, compared with the previous state of the art [14]. We expect the key insight of the paper (i.e., explicitly handling domain difference via domain transport) will prove to be useful for other similar problems in computer vision as well.

Lending Orientation to Neural Networks for Cross-view Geo-localization

This paper studies image-based geo-localization (IBL) problem using ground-to-aerial cross-view matching. The goal is to predict the spatial location of a ground-level query image by matching it to a large geotagged aerial image database (e.g., satellite imagery). This is a challenging task due to the drastic differences in their viewpoints and visual appearances. Existing deep learning methods for this problem have been focused on maximizing feature similarity between spatially close-by image pairs, while minimizing other images pairs which are far apart. They do so by deep feature embedding based on visual appearance in those ground-and-aerial images. However, in everyday life, humans commonly use {\em orientation} information as an important cue for the task of spatial localization. Inspired by this insight, this paper proposes a novel method which endows deep neural networks with the `commonsense' of orientation. Given a ground-level spherical panoramic image as query input (and a large georeferenced satellite image database), we design a Siamese network which explicitly encodes the orientation (i.e., spherical directions) of each pixel of the images. Our method significantly boosts the discriminative power of the learned deep features, leading to a much higher recall and precision outperforming all previous methods. Our network is also more compact using only 1/5th number of parameters than a previously best-performing network. To evaluate the generalization of our method, we also created a large-scale cross-view localization benchmark containing 100K geotagged ground-aerial pairs covering a city. Our codes and datasets are available at \url{https://github.com/Liumouliu/OriCNN}.

High Dimensional Robust Estimation of Sparse Models via Trimmed Hard Thresholding

We study the problem of sparsity constrained $M$-estimation with arbitrary corruptions to both {\em explanatory and response} variables in the high-dimensional regime, where the number of variables $d$ is larger than the sample size $n$. Our main contribution is a highly efficient gradient-based optimization algorithm that we call Trimmed Hard Thresholding -- a robust variant of Iterative Hard Thresholding (IHT) by using trimmed mean in gradient computations. Our algorithm can deal with a wide class of sparsity constrained $M$-estimation problems, and we can tolerate a nearly dimension independent fraction of arbitrarily corrupted samples. More specifically, when the corrupted fraction satisfies $\epsilon \lesssim {1} /\left({\sqrt{k} \log (nd)}\right)$, where $k$ is the sparsity of the parameter, we obtain accurate estimation and model selection guarantees with optimal sample complexity. Furthermore, we extend our algorithm to sparse Gaussian graphical model (precision matrix) estimation via a neighborhood selection approach. We demonstrate the effectiveness of robust estimation in sparse linear, logistic regression, and sparse precision matrix estimation on synthetic and real-world US equities data.

Dynamic Sparse Graph for Efficient Deep Learning

We propose to execute deep neural networks (DNNs) with dynamic and sparse graph (DSG) structure for compressive memory and accelerative execution during both training and inference. The great success of DNNs motivates the pursuing of lightweight models for the deployment onto embedded devices. However, most of the previous studies optimize for inference while neglect training or even complicate it. Training is far more intractable, since (i) the neurons dominate the memory cost rather than the weights in inference; (ii) the dynamic activation makes previous sparse acceleration via one-off optimization on fixed weight invalid; (iii) batch normalization (BN) is critical for maintaining accuracy while its activation reorganization damages the sparsity. To address these issues, DSG activates only a small amount of neurons with high selectivity at each iteration via a dimension-reduction search (DRS) and obtains the BN compatibility via a double-mask selection (DMS). Experiments show significant memory saving (1.7-4.5x) and operation reduction (2.3-4.4x) with little accuracy loss on various benchmarks.