Image Visual Sentiment Analysis (VSA) requires the availability of large annotated datasets, whose construction presents many challenges. The necessity of gathering a large amount of labeled images contrasts with the rigorous, but lengthy, process required for manual annotation based on psychovisual experiments, and with the automatic gathering of large amounts of data roughly labeled based on the sentiment analysis of the text accompanying the images, like captions, tweets and tags. An additional limitation is the scarcity of high-quality datasets with a neutral class, which forces the images to be classified into emotions even when the observers show no emotional activation. In this work, we present a scalable methodology rooted in semiotics and art theory for the construction of a 3-class (positive, negative and neutral) VSA dataset, enabling the downloading of a desired quantity of images while maintaining labeling coherence and accuracy. Based on the proposed methodology, we introduce and make publicly available a VSA dataset of over 100,000 images. To validate the quality of the dataset, we used it to train several classifiers and compared their performance with those of classifiers trained on other datasets. The results, we got, show that the classifiers trained on the new dataset provide better performance when tested on independent datasets, including those commonly used for psycho-visual experiments.

A common approach to improve the robustness of synthetic image detectors against image post-processing is to augment the dataset the detectors are trained on by applying a selected pool of image processing operators. A list of commonly adopted image processing augmentations includes JPEG compression, geometric transformations, color adjustment, noise addition, and filtering. Robustness against image processing operators that are not included in the augmentation pool, however, is problematic since the detectors tend to overfit to the image operators used during training, without generalizing to other kinds of processing. In this paper, we introduce a new form of data augmentation based on the simulation of the Print & Scan (P&S) process. We argue that asking the synthetic image detector to still work after that an image has been printed and scanned, forces the detector to rely on robust features that can be detected even after other forms of processing. Given the impossibility of creating a large enough dataset of P&S images, we trained a CycleGAN network to simulate the P&S process and used it for data augmentation. The results we got by applying the above procedure to a detector trained to distinguish real and synthetic images in different domains show that P&S augmentation improves the robustness of the detectors even on images processed by operators that have not been used during training.

Despite the wide variety of methods developed for synthetic image attribution, most of them can only attribute images generated by models or architectures included in the training set and do not work with unknown architectures, hindering their applicability in real-world scenarios. In this paper, we propose a verification framework that relies on a Siamese Network to address the problem of open-set attribution of synthetic images to the architecture that generated them. We consider two different settings. In the first setting, the system determines whether two images have been produced by the same generative architecture or not. In the second setting, the system verifies a claim about the architecture used to generate a synthetic image, utilizing one or multiple reference images generated by the claimed architecture. The main strength of the proposed system is its ability to operate in both closed and open-set scenarios so that the input images, either the query and reference images, can belong to the architectures considered during training or not. Experimental evaluations encompassing various generative architectures such as GANs, diffusion models, and transformers, focusing on synthetic face image generation, confirm the excellent performance of our method in both closed and open-set settings, as well as its strong generalization capabilities.

Owing to their realistic features and continuous improvements, images manipulated by Generative Adversarial Network (GAN) have become a compelling research topic.In this paper, we apply detection and localization to GAN manipulated images by means of models, based on EfficientNet-B4 architectures.Detection is tested on multiple generated multi-spectral datasets from several world regions and different GAN architectures, whereas localization is tested on an inpainted images dataset of sizes 2048×2048×13.The results obtained for both detection and localization are shown to be promising.

A new approach based on a mixture of Gaussian and quadrilateral functions was developed to process bathymetric lidar waveforms. The approach was tested on two simulated data sets obtained from the existing Water-LIDAR (Wa-LID) waveform simulator. The first simulated data set corresponds to a sensor configuration modeled after a possible future satellite bathymetric lidar sensor that was previously studied. The second simulated data set corresponds to a lidar airborne configuration modeled using the HawkEye airborne lidar parameters. In the proposed approach, the lidar waveform is fitted into a combination of three functions, two Gaussians for both the water surface and water bottom contributions and a quadrilateral function to fit the water column contribution. The results show more accurate bathymetry estimates compared with the use of a triangular function to fit the column contribution or a simple peak detection method. For the satellite configuration, the bias is improved by 16.8 and 0.8 cm compared with the peak detection method and the use of a triangular function, respectively. For the airborne configuration, the bias is improved by 10.0 and 2.4 cm compared with the peak detection method and the use of a triangular function, respectively. The proposed waveform fitting using the quadrilateral function underestimates the bathymetry by$-$5.0 and$-$6.1 cm for the simulated satellite and airborne data sets, respectively. The standard deviations of the bathymetry estimates are 6.0 and 8.2 cm, respectively. The obtained biases are inherent to overlaps between functions fitting the water surface, column, and bottom contributions.

Bathymetry is usually determined using the positions of the water surface and the water bottom peaks of the green LiDAR waveform. The water bottom peak characteristics are known to be sensitive to the bottom slope, which induces pulse stretching. However, the effects of a more complex bottom geometry within the footprint below semitransparent media are less understood. In this letter, the effects of the water bottom geometry on the shifting of the bottom peaks in the waveforms were modeled. For the sake of simplicity, the bottom geometry is modeled as a 1D sequence of successive contiguous segments with various slopes. The positions of the peaks in waveforms were deduced using a conventional peak detection process on simulated waveforms. The waveforms were simulated using the existing Wa-LID waveform simulator, which was extended in this study to account for a 1D complex bottom geometry. An experimental design using various water depths, bottom slopes, and LiDAR footprint sizes according to the design of satellite sensors was used for the waveform simulation. Power laws that explained the peak time shifting as a function of the footprint size and the water bottom slope were approximated. Peak shifting induces a bias in the bathymetry estimates that is based on a peak detection of up to 92% of the true water depth. This bias may also explain the frequent underestimation of the water depth from bathymetric airborne LiDAR surveys observed in various empirical studies.