The effectiveness of GANs in producing images according to a specific visual domain has shown potential in unsupervised domain adaptation. Source labeled images have been modified to mimic target samples for training classifiers in the target domain, and inverse mappings from the target to the source domain have also been evaluated, without new image generation. In this paper we aim at getting the best of both worlds by introducing a symmetric mapping among domains. We jointly optimize bi-directional image transformations combining them with target self-labeling. We define a new class consistency loss that aligns the generators in the two directions, imposing to preserve the class identity of an image passing through both domain mappings. A detailed analysis of the reconstructed images, a thorough ablation study and extensive experiments on six different settings confirm the power of our approach.

Vision Transformer (ViT) architectures are becoming increasingly popular and widely employed to tackle computer vision applications. Their main feature is the capacity to extract global information through the self-attention mechanism, outperforming earlier convolutional neural networks. However, ViT deployment and performance have grown steadily with their size, number of trainable parameters, and operations. Furthermore, self-attention's computational and memory cost quadratically increases with the image resolution. Generally speaking, it is challenging to employ these architectures in real-world applications due to many hardware and environmental restrictions, such as processing and computational capabilities. Therefore, this survey investigates the most efficient methodologies to ensure sub-optimal estimation performances. More in detail, four efficient categories will be analyzed: compact architecture, pruning, knowledge distillation, and quantization strategies. Moreover, a new metric called Efficient Error Rate has been introduced in order to normalize and compare models’ features that affect hardware devices at inference time, such as the number of parameters, bits, FLOPs, and model size. Summarizing, this paper first mathematically defines the strategies used to make Vision Transformer efficient, describes and discusses state-of-the-art methodologies, and analyzes their performances over different application scenarios. Toward the end of this paper, we also discuss open challenges and promising research directions.

Depth estimation is a fundamental knowledge for autonomous systems that need to assess their own state and perceive the surrounding environment. Deep learning algorithms for depth estimation have gained significant interest in recent years, owing to the potential benefits of this methodology in overcoming the limitations of active depth sensing systems. Moreover, due to the low cost and size of monocular cameras, researchers have focused their attention on monocular depth estimation (MDE), which consists in estimating a dense depth map from a single RGB video frame. State of the art MDE models typically rely on vision transformers (ViT) architectures that are highly deep and complex, making them unsuitable for fast inference on devices with hardware constraints. Purposely, in this paper, we address the problem of exploiting ViT in MDE on embedded devices. Those systems are usually characterized by limited memory capabilities and low-power CPU/GPU. We propose METER, a novel lightweight vision transformer architecture capable of achieving state of the art estimations and low latency inference performances on the considered embedded hardwares: NVIDIA Jetson TX1 and NVIDIA Jetson Nano. We provide a solution consisting of three alternative configurations of METER, a novel loss function to balance pixel estimation and reconstruction of image details, and a new data augmentation strategy to improve the overall final predictions. The proposed method outperforms previous lightweight works over the two benchmark datasets: the indoor NYU Depth v2 and the outdoor KITTI.

The ability to classify objects is fundamental for robots. Besides knowledge about their visual appearance, captured by the RGB channel, robots heavily need also depth information to make sense of the world. While the use of deep networks on RGB robot images has benefited from the plethora of results obtained on databases like ImageNet, using convnets on depth images requires mapping them into three-dimensional channels. This transfer learning procedure makes them processable by pretrained deep architectures. Current mappings are based on heuristic assumptions over preprocessing steps and on what depth properties should be most preserved, resulting often in cumbersome data visualizations, and in suboptimal performance in terms of generality and recognition results. Here, we take an alternative route and we attempt instead to learn an optimal colorization mapping for any given pretrained architecture, using as training data a reference RGB-D database. We propose a deep network architecture, exploiting the residual paradigm, that learns how to map depth data to three channel images. A qualitative analysis of the images obtained with this approach clearly indicates that learning the optimal mapping preserves the richness of depth information better than current hand-crafted approaches. Experiments on the Washington, JHUIT-50 and BigBIRD public benchmark databases, using CaffeNet, VGG-16, GoogleNet, and ResNet50 clearly showcase the power of our approach, with gains in performance of up to 16% compared to state of the art competitors on the depth channel only, leading to top performances when dealing with RGB-D data.

Raman spectroscopy (RS) is a label-free molecular vibrational spectroscopy technique that is able to identify the molecular fingerprint of various samples making use of the inelastic scattering of monochromatic light. Because of its advantages of non-destructive and accurate detection, RS is finding more and more use for benign and malignant tissues, tumor differentiation, tumor subtype classification, and section pathology diagnosis, operating either in vivo or in vitro. However, the high specificity of RS comes at a cost. The acquisition rate is low, depth information cannot be directly accessed, and the sampling area is limited. Such limitations can be contained if data pre- and post-processing methods are combined with current methods of Artificial Intelligence (AI), essentially, Machine Learning (ML) and Deep Learning (DL). The latter is modifying the approach to cancer diagnosis currently used to automate many cancer data analyses, and it has emerged as a promising option for improving healthcare accuracy and patient outcomes by abiliting prediction diseases tools. In a very broad context, Artificial Intelligence applications in oncology include risk assessment, early diagnosis, patient prognosis estimation, and treatment selection based on deep knowledge. The application of autonomous methods to datasets generated by RS analysis of benign and malignant tissues could make RS a rapid and stand-alone technique to help pathologists diagnose cancer with very high accuracy. This review describes the current milestones achieved by applying AI-based algorithms to RS analysis, grouped according to seven major types of cancers (Pancreatic, Breast, Skin, Brain, Prostate, Ovarian and Oral cavity). Additionally, it provides a theoretical foundation to tackle both present and forthcoming challenges in this domain. By exploring the current achievements and discussing the relative methodologies, this review offers recapitulative insights on recent and ongoing efforts to position RS as a rapid and effective cancer screening tool for pathologists. Accordingly, we aim to encourage future research endeavors and to facilitate the realization of the full potential of RS and AI applications in cancer grading.

For the purpose of maintaining and prolonging the service life of civil constructions, structural damage must be closely monitored. Monitoring the incidence, formation, and spread of damage is crucial to ensure a structure’s ongoing performance. This research proposes a unique approach for multiclass damage detection using acceleration responses based on synchrosqueezing transform (SST) together with deep learning algorithms. In particular, our pipeline is able to classify correctly the time series representing the responses of accelerometers placed on a bridge, which are classified with respect to different types of damage scenarios applied to the bridge. Using benchmark data from the Z24 bridge for multiclass classification for different damage situations, the suggested method is validated. This dataset includes labeled accelerometer measurements from a real-world bridge that has been gradually damaged by various conditions. The findings demonstrate that the suggested approach is successful in exploiting pre-trained 2D convolutional neural networks, obtaining a high classification accuracy that can be further boosted by the application of simple voting methods.