Image forensics is an increasingly relevant problem, as it can potentially address online disinformation campaigns and mitigate problematic aspects of social media. Of particular interest, given its recent successes, is the detection of imagery produced by Generative Adversarial Networks (GANs), e.g. ‘deepfakes’. Leveraging large training sets and extensive computing resources, recent GANs can be trained to generate synthetic imagery which is (in some ways) indistinguishable from real imagery. We analyze the structure of the generating network of a popular GAN implementation [1], and show that the network’s treatment of exposure is markedly different from a real camera. We further show that this cue can be used to distinguish GAN-generated imagery from camera imagery, including effective discrimination between GAN imagery and real camera images used to train the GAN.

We match three hadronic equations of state at low energy densities to a perturbatively computed equation of state of quarks and gluons at high energy densities. One of them includes all known hadrons treated as point particles, which approximates attractive interactions among hadrons. The other two include, in addition, repulsive interactions in the form of excluded volumes occupied by the hadrons. A switching function is employed to make the crossover transition from one phase to another without introducing a thermodynamic phase transition. A ${\chi }^2$ fit to accurate lattice calculations with temperature $100<T<1000$ MeV determines the parameters. These parameters quantify the behavior of the QCD running gauge coupling and the hard core radius of protons and neutrons, which turns out to be $0.62\pm 0.04$ fm. The most physically reasonable models include the excluded-volume effect. Not only do they include the effects of attractive and repulsive interactions among hadrons, but they also achieve better agreement with lattice QCD calculations of the equation of state. The equations of state constructed in this paper do not result in a phase transition, at least not for the temperatures and baryon chemical potentials investigated. It remains to be seen how well these equations of state will represent experimental data on high-energy heavy-ion collisions when implemented in hydrodynamic simulations.

We develop a flexible quasiparticle theory of transport coefficients of hot hadronic matter at finite baryon density. We begin with a hadronic quasiparticle model which includes a scalar and a vector mean field. Quasiparticle energies and the mean fields depend on temperature and baryon chemical potential. Starting with the quasiparticle dispersion relation, we derive the Boltzmann equation and use the Chapman-Enskog expansion to derive formulas for the shear and bulk viscosities and thermal conductivity. We obtain both relaxation-time approximation formulas and more general integral equations. Throughout the work, we explicitly enforce the Landau-Lifshitz conditions of fit and ensure the theory is thermodynamically self-consistent. The derived formulas should be useful for predicting the transport coefficients of the hadronic phase of matter produced in heavy-ion collisions at the Relativistic Heavy Ion Collider and at other accelerators.

Fluctuations of the proton number distribution in central Au-Au collisions have been measured by the STAR Collaboration in a beam energy scan at the Relativistic Heavy Ion Collider (RHIC). The motivation is a search for evidence of a critical point in the equation of state. It was found that the skewness and kurtosis display an interesting energy dependence. We compare these measurements to an equation of state which smoothly interpolates between an excluded volume hadron resonance gas at low energy density to a perturbative plasma of quarks and gluons at high energy density. This crossover equation of state agrees very well with the lattice QCD equation of state. The crossover equation of state can reproduce the data if the fluctuations are frozen at a temperature significantly lower than the average chemical freeze-out.

What is the current state-of-the-art for image restoration and enhancement applied to degraded images acquired under less than ideal circumstances? Can the application of such algorithms as a pre-processing step improve image interpretability for manual analysis or automatic visual recognition to classify scene content? While there have been important advances in the area of computational photography to restore or enhance the visual quality of an image, the capabilities of such techniques have not always translated in a useful way to visual recognition tasks. Consequently, there is a pressing need for the development of algorithms that are designed for the joint problem of improving visual appearance and recognition, which will be an enabling factor for the deployment of visual recognition tools in many real-world scenarios. To address this, we introduce the UG $^2$ dataset as a large-scale benchmark composed of video imagery captured under challenging conditions, and two enhancement tasks designed to test algorithmic impact on visual quality and automatic object recognition. Furthermore, we propose a set of metrics to evaluate the joint improvement of such tasks as well as individual algorithmic advances, including a novel psychophysics-based evaluation regime for human assessment and a realistic set of quantitative measures for object recognition performance. We introduce six new algorithms for image restoration or enhancement, which were created as part of the IARPA sponsored UG $^2$ Challenge workshop held at CVPR 2018. Under the proposed evaluation regime, we present an in-depth analysis of these algorithms and a host of deep learning-based and classic baseline approaches. From the observed results, it is evident that we are in the early days of building a bridge between computational photography and visual recognition, leaving many opportunities for innovation in this area.