arXiv subjects

Cong Cao

Publications and source records attributed to Cong Cao.

47 records · Page 3Linked to original sources

arXiv2023

Unsupervised HDR Image and Video Tone Mapping via Contrastive Learning

Capturing high dynamic range (HDR) images (videos) is attractive because it can reveal the details in both dark and bright regions. Since the mainstream screens only support low dynamic range (LDR) content, tone mapping algorithm is required to compress the dynamic range of HDR images (videos). Although image tone mapping has been widely explored, video tone mapping is lagging behind, especially for the deep-learning-based methods, due to the lack of HDR-LDR video pairs. In this work, we propose a unified framework (IVTMNet) for unsupervised image and video tone mapping. To improve unsupervised training, we propose domain and instance based contrastive learning loss. Instead of using a universal feature extractor, such as VGG to extract the features for similarity measurement, we propose a novel latent code, which is an aggregation of the brightness and contrast of extracted features, to measure the similarity of different pairs. We totally construct two negative pairs and three positive pairs to constrain the latent codes of tone mapped results. For the network structure, we propose a spatial-feature-enhanced (SFE) module to enable information exchange and transformation of nonlocal regions. For video tone mapping, we propose a temporal-feature-replaced (TFR) module to efficiently utilize the temporal correlation and improve the temporal consistency of video tone-mapped results. We construct a large-scale unpaired HDR-LDR video dataset to facilitate the unsupervised training process for video tone mapping. Experimental results demonstrate that our method outperforms state-of-the-art image and video tone mapping methods. Our code and dataset are available at https://github.com/cao-cong/UnCLTMO.

cs.CV↗

arXiv2022

Adversarial Dual-Student with Differentiable Spatial Warping for Semi-Supervised Semantic Segmentation

A common challenge posed to robust semantic segmentation is the expensive data annotation cost. Existing semi-supervised solutions show great potential for solving this problem. Their key idea is constructing consistency regularization with unsupervised data augmentation from unlabeled data for model training. The perturbations for unlabeled data enable the consistency training loss, which benefits semi-supervised semantic segmentation. However, these perturbations destroy image context and introduce unnatural boundaries, which is harmful for semantic segmentation. Besides, the widely adopted semi-supervised learning framework, i.e. mean-teacher, suffers performance limitation since the student model finally converges to the teacher model. In this paper, first of all, we propose a context friendly differentiable geometric warping to conduct unsupervised data augmentation; secondly, a novel adversarial dual-student framework is proposed to improve the Mean-Teacher from the following two aspects: (1) dual student models are learned independently except for a stabilization constraint to encourage exploiting model diversities; (2) adversarial training scheme is applied to both students and the discriminators are resorted to distinguish reliable pseudo-label of unlabeled data for self-training. Effectiveness is validated via extensive experiments on PASCAL VOC2012 and Cityscapes. Our solution significantly improves the performance and state-of-the-art results are achieved on both datasets. Remarkably, compared with fully supervision, our solution achieves comparable mIoU of 73.4% using only 12.5% annotated data on PASCAL VOC2012. Our codes and models are available at https://github.com/cao-cong/ADS-SemiSeg.

cs.CV↗

arXiv2022

PHN: Parallel heterogeneous network with soft gating for CTR prediction

The Click-though Rate (CTR) prediction task is a basic task in recommendation system. Most of the previous researches of CTR models built based on Wide \& deep structure and gradually evolved into parallel structures with different modules. However, the simple accumulation of parallel structures can lead to higher structural complexity and longer training time. Based on the Sigmoid activation function of output layer, the linear addition activation value of parallel structures in the training process is easy to make the samples fall into the weak gradient interval, resulting in the phenomenon of weak gradient, and reducing the effectiveness of training. To this end, this paper proposes a Parallel Heterogeneous Network (PHN) model, which constructs a network with parallel structure through three different interaction analysis methods, and uses Soft Selection Gating (SSG) to feature heterogeneous data with different structure. Finally, residual link with trainable parameters are used in the network to mitigate the influence of weak gradient phenomenon. Furthermore, we demonstrate the effectiveness of PHN in a large number of comparative experiments, and visualize the performance of the model in training process and structure.

cs.LG↗

arXiv2020

Rheology finds distinct glass and jamming transitions in emulsions

We study the rheology of monodisperse and bidisperse emulsions with various droplet sizes (1 $μ$m -- 2 $μ$m diameter). Above a critical volume fraction $ϕ_c$, these systems exhibit solid-like behavior and a yield stress can be detected. Previous experiments suggest that for small thermal particles, rheology will see a glass transition at $ϕ_c = ϕ_g =0.58$; for large athermal systems, rheology will see a jamming transition at $ϕ_c = ϕ_J =0.64$. However, simulations point out that at the crossover of thermal and athermal regimes, the glass and jamming transitions may both be observed in the same sample. Here we conduct an experiment by shearing four oil-in-water emulsions with a rheometer. We observe both a glass and a jamming transition for our smaller diameter droplets, and only a jamming transition for our larger diameter droplets. The bidisperse sample behaves similarly to the small droplet sample, with two transitions observed. Our rheology data are well-fit by both the Herschel-Bulkley model and the Three Component model. Based on the fitting parameters, our raw rheological data would not collapse onto a master curve. Our results suggest that liquid-solid transitions in dispersions may not be universal, but depend on particle type.

cond-mat.soft↗

arXiv2020

Supervised Raw Video Denoising with a Benchmark Dataset on Dynamic Scenes

In recent years, the supervised learning strategy for real noisy image denoising has been emerging and has achieved promising results. In contrast, realistic noise removal for raw noisy videos is rarely studied due to the lack of noisy-clean pairs for dynamic scenes. Clean video frames for dynamic scenes cannot be captured with a long-exposure shutter or averaging multi-shots as was done for static images. In this paper, we solve this problem by creating motions for controllable objects, such as toys, and capturing each static moment for multiple times to generate clean video frames. In this way, we construct a dataset with 55 groups of noisy-clean videos with ISO values ranging from 1600 to 25600. To our knowledge, this is the first dynamic video dataset with noisy-clean pairs. Correspondingly, we propose a raw video denoising network (RViDeNet) by exploring the temporal, spatial, and channel correlations of video frames. Since the raw video has Bayer patterns, we pack it into four sub-sequences, i.e RGBG sequences, which are denoised by the proposed RViDeNet separately and finally fused into a clean video. In addition, our network not only outputs a raw denoising result, but also the sRGB result by going through an image signal processing (ISP) module, which enables users to generate the sRGB result with their favourite ISPs. Experimental results demonstrate that our method outperforms state-of-the-art video and raw image denoising algorithms on both indoor and outdoor videos.

eess.IV↗

arXiv2019

Returning Scientists and the Emergence of Chinese Science System

The Chinese approach to developing a world-class science system includes a vigorous set of programmes to attract back Chinese researchers who have overseas training and work experience. No analysis is available to show the performance of these mobile researchers. This article attempts to close part of this gap. Using a novel bibliometric approach, we estimate the stocks of overseas Chinese and returnees from the perspective of their publication activities, albeit with some limitations. We show that the share of overseas Chinese scientists in the US is considerably larger than that in the EU. We also show that Chinese returnees publish higher impact work, and continue to publish more and at the international level than domestic counterparts. Returnees not only tend to publish more, but they are instrumental in linking China into the global network. Indeed, returnees actively co-publish with researchers in their former host system, showing the importance of scientific social capital. Future research will examine the impact of length of stay, among other factors, on such impact and integration.

stat.AP↗

arXiv2019

Random packing of rods in small containers

We conduct experiments and simulations to study the disordered packing of rods in small containers. Experiments study cylindrical rods with aspect ratio ranging from 4 to 32; simulations use of spherocylinders with similar aspect ratios. In all cases, rods pack randomly in cylindrical containers whose smallest dimension is larger than the rod length. Packings in smaller containers have lower volume fractions than those in larger containers, demonstrating the influence of the boundaries. The volume fraction extrapolated to infinite container size decreases with increasing aspect ratio, in agreement with previous work. X-ray tomography experiments show that the boundary effects depend on the orientation of the boundary, indicating a strong influence of gravity, whereas the simulation finds boundary effects that are purely geometric. In all cases, the boundary influence extends approximately half a particle length into the interior of the container.

cond-mat.soft↗

arXiv2018

The Behrens-Fisher Problem with Covariates and Baseline Adjustments

The Welch-Satterthwaite t-test is one of the most prominent and often used statistical inference method in applications. The method is, however, not flexible with respect to adjustments for baseline values or other covariates, which may impact the response variable. Existing analysis of covariance methods are typically based on the assumption of equal variances across the groups. This assumption is hard to justify in real data applications and the methods tend to not control the type-1 error rate satisfactorily under variance heteroscedasticity. In the present paper, we tackle this problem and develop unbiased variance estimators of group specific variances, and especially of the variance of the estimated adjusted treatment effect in a general analysis of covariance model. These results are used to generalize the Welch-Satterthwaite t-test to covariates adjustments. Extensive simulation studies show that the method accurately controls the nominal type-1 error rate, even for very small sample sizes, moderately skewed distributions and under variance heteroscedasticity. A real data set motivates and illustrates the application of the proposed methods.

stat.ME↗

arXiv2017

Aging near rough and smooth boundaries in colloidal glasses

We use confocal microscopy to study the aging of a bidisperse colloidal glass near rough and smooth boundaries. Near smooth boundaries, the particles form layers, and particle motion is dramatically slower near the boundary as compared to the bulk. Near rough boundaries, the layers nearly vanish, and particle motion is nearly identical to that of the bulk. The gradient in dynamics near the boundaries is demonstrated to be a function of the gradient in structure for both types of boundaries.Our observations show that wall-induced layer structures strongly influence aging.

cond-mat.soft↗

arXiv2017

Cavity mediated effective interaction between phonon and magnon

Optomagnonics and optomechanics have various applications ranging from tunable light sources to optical manipulation for quantum information science. Here, we propose a hybrid system with the interaction between phonon and magnon which could be tuned by the electromagnetic field based on the radiation pressure and magneto-optical effect. The self energy of the magnon and phonon induced by the electromagnetic field and the influence of the thermal noise are studied. Moreover, the topological features of the hybrid system is illustrated with resort to the dynamical encircling with the exceptional points, and the chiral characteristics under these encirclements are found.

quant-ph↗

arXiv2016

Effective mass sensing using optomechanically induced transparency in microresonator system

Detecting and weighing the individual nanoparticles is an important approach to study the behavior and properties of single particles. Here we illustrate an effective mass sensing scheme using optomechanical resonator system. Based on the optomechanically induced transparency phenomenon, a Stokes field reference approach is used to sense the mass of the particle on the microresonator. The field intensity of the transmission field will be changed by the effect of the particle which avoids the limits of decay induced spectral width in the resonance shift detection. Exploiting the perturbation method, we theoretically evaluated the dynamical behavior of the system, and achieved the femtogram level mass sensing without the need for high cavity Q-value, and strong coupling strength in the optomechanically system.

quant-ph↗