Paolo Russo

Lorenzo Papa, Paolo Russo, Irene Amerini et al.

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 firstly 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.

L. Papa, P. Russo, I. Amerini

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.

L. Papa, P. Russo, I. Amerini

Ground-truth RGBD data are fundamental for a wide range of computer vision applications; however, those labeled samples are difficult to collect and time-consuming to produce. A common solution to overcome this lack of data is to employ graphic engines to produce synthetic proxies; however, those data do not often reflect real-world images, resulting in poor performance of the trained models at the inference step. In this paper we propose a novel training pipeline that incorporates Diffusion4D (D4D), a customized 4-channels diffusion model able to generate realistic RGBD samples. We show the effectiveness of the developed solution in improving the performances of deep learning models on the monocular depth estimation task, where the correspondence between RGB and depth map is crucial to achieving accurate measurements. Our supervised training pipeline, enriched by the generated samples, outperforms synthetic and original data performances achieving an RMSE reduction of (8.2%, 11.9%) and (8.1%, 6.1%) respectively on the indoor NYU Depth v2 and the outdoor KITTI dataset.

Paolo Russo, Fabiana Di Ciaccio, Salvatore Troisi

One of the main issues for underwater robots navigation is their accurate positioning, which heavily depends on the orientation estimation phase. The systems employed to this scope are affected by different noise typologies, mainly related to the sensors and to the irregular noise of the underwater environment. Filtering algorithms can reduce their effect if opportunely configured, but this process usually requires fine techniques and time. In this paper we propose DANAE, a deep Denoising AutoeNcoder for Attitude Estimation which works on Kalman filter IMU/AHRS data integration with the aim of reducing any kind of noise, independently of its nature. This deep learning-based architecture showed to be robust and reliable, significantly improving the Kalman filter results. Further tests could make this method suitable for real-time applications on navigation tasks.

Fabio M. Carlucci, Paolo Russo, Tatiana Tommasi et al.

The ability to generalize across visual domains is crucial for the robustness of artificial recognition systems. Although many training sources may be available in real contexts, the access to even unlabeled target samples cannot be taken for granted, which makes standard unsupervised domain adaptation methods inapplicable in the wild. In this work we investigate how to exploit multiple sources by hallucinating a deep visual domain composed of images, possibly unrealistic, able to maintain categorical knowledge while discarding specific source styles. The produced agnostic images are the result of a deep architecture that applies pixel adaptation on the original source data guided by two adversarial domain classifier branches at image and feature level. Our approach is conceived to learn only from source data, but it seamlessly extends to the use of unlabeled target samples. Remarkable results for both multi-source domain adaptation and domain generalization support the power of hallucinating agnostic images in this framework.

Paolo Russo, Fabio Maria Carlucci, Tatiana Tommasi et al.

The effectiveness of generative adversarial approaches in producing images according to a specific style or visual domain has recently opened new directions to solve the unsupervised domain adaptation problem. It has been shown that source labeled images can be modified to mimic target samples making it possible to train directly a classifier in the target domain, despite the original lack of annotated data. Inverse mappings from the target to the source domain have also been evaluated but only passing through adapted feature spaces, thus without new image generation. In this paper we propose to better exploit the potential of generative adversarial networks for adaptation by introducing a novel symmetric mapping among domains. We jointly optimize bi-directional image transformations combining them with target self-labeling. Moreover we define a new class consistency loss that aligns the generators in the two directions imposing to conserve the class identity of an image passing through both domain mappings. A detailed qualitative and quantitative analysis of the reconstructed images confirm the power of our approach. By integrating the two domain specific classifiers obtained with our bi-directional network we exceed previous state-of-the-art unsupervised adaptation results on four different benchmark datasets.

F. M. Carlucci, P. Russo, B. Caputo

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 pre-trained 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 sub-optimal 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 pre-trained 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, VGG16, 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

Fabio Maria Carlucci, Paolo Russo, Barbara Caputo

Convolutional Neural Networks (CNNs) trained on large scale RGB databases have become the secret sauce in the majority of recent approaches for object categorization from RGB-D data. Thanks to colorization techniques, these methods exploit the filters learned from 2D images to extract meaningful representations in 2.5D. Still, the perceptual signature of these two kind of images is very different, with the first usually strongly characterized by textures, and the second mostly by silhouettes of objects. Ideally, one would like to have two CNNs, one for RGB and one for depth, each trained on a suitable data collection, able to capture the perceptual properties of each channel for the task at hand. This has not been possible so far, due to the lack of a suitable depth database. This paper addresses this issue, proposing to opt for synthetically generated images rather than collecting by hand a 2.5D large scale database. While being clearly a proxy for real data, synthetic images allow to trade quality for quantity, making it possible to generate a virtually infinite amount of data. We show that the filters learned from such data collection, using the very same architecture typically used on visual data, learns very different filters, resulting in depth features (a) able to better characterize the different facets of depth images, and (b) complementary with respect to those derived from CNNs pre-trained on 2D datasets. Experiments on two publicly available databases show the power of our approach.

Tatiana Tommasi, Martina Lanzi, Paolo Russo et al.

In this paper we focus on the spatial nature of visual domain shift, attempting to learn where domain adaptation originates in each given image of the source and target set. We borrow concepts and techniques from the CNN visualization literature, and learn domainnes maps able to localize the degree of domain specificity in images. We derive from these maps features related to different domainnes levels, and we show that by considering them as a preprocessing step for a domain adaptation algorithm, the final classification performance is strongly improved. Combined with the whole image representation, these features provide state of the art results on the Office dataset.