Abstract
It is well known that the performance of Machine Learning techniques, notably when applied to Computer Vision (CV), depends heavily on the amount and quality of the training data set. However, large data sets lead to time-consuming training loops and, in many situations, are difficult or even impossible to create. Therefore, there is a need for solutions to reduce their size while ensuring good levels of performance, i.e., solutions that obtain the best tradeoff between the amount/quality of training data and the model's performance. This paper proposes a dataset reduction approach for training data used in Deep Learning methods in Facial Recognition (FR) problems. We focus on maximizing the variability of representations for each subject (person) in the training data, thus favoring quality instead of size. The main research questions are: 1) Which facial features better discriminate different identities? 2) Will it be possible to significantly reduce the training time without compromising performance? 3) Should we favor quality over quantity for very large datasets in FR? This analysis uses a pipeline to discriminate a set of features suitable for capturing the diversity and a cluster-based sampling to select the best images for each training subject, i.e., person. Results were obtained using VGGFace2 and Labeled Faces in the Wild (for benchmarking) and show that, with the proposed approach, a data reduction is possible while ensuring similar levels of accuracy.

Abstract
We propose a unified architecture for visual scene understanding, aimed at overcoming the limitations of traditional, fragmented approaches in computer vision. Our work focuses on creating a system that accurately and coherently interprets visual scenes, with the ultimate goal to provide a 3D virtual representation, which is particularly useful for applications in virtual and augmented reality. By integrating various visual and semantic processing tasks into a single, adaptable framework, our architecture simplifies the design process, ensuring a seamless and consistent scene interpretation. This is particularly important in complex systems that rely on 3D synthesis, as the need for precise and semantically coherent scene descriptions keeps on growing. Our unified approach addresses these challenges, offering a flexible and efficient solution. We demonstrate the practical effectiveness of our architecture through a proof-of-concept system and explore its potential in various application domains, proving its value in advancing the field of computer vision.

Abstract
Purpose: Visual perception enables robots to perceive the environment. Visual data is processed using computer vision algorithms that are usually time-expensive and require powerful devices to process the visual data in real-time, which is unfeasible for open-field robots with limited energy. This work benchmarks the performance of different heterogeneous platforms for object detection in real-time. This research benchmarks three architectures: embedded GPU-Graphical Processing Units (such as NVIDIA Jetson Nano 2 GB and 4 GB, and NVIDIA Jetson TX2), TPU-Tensor Processing Unit (such as Coral Dev Board TPU), and DPU-Deep Learning Processor Unit (such as in AMD-Xilinx ZCU104 Development Board, and AMD-Xilinx Kria KV260 Starter Kit). Methods: The authors used the RetinaNet ResNet-50 fine-tuned using the natural VineSet dataset. After the trained model was converted and compiled for target-specific hardware formats to improve the execution efficiency.Conclusions and Results: The platforms were assessed in terms of performance of the evaluation metrics and efficiency (time of inference). Graphical Processing Units (GPUs) were the slowest devices, running at 3 FPS to 5 FPS, and Field Programmable Gate Arrays (FPGAs) were the fastest devices, running at 14 FPS to 25 FPS. The efficiency of the Tensor Processing Unit (TPU) is irrelevant and similar to NVIDIA Jetson TX2. TPU and GPU are the most power-efficient, consuming about 5 W. The performance differences, in the evaluation metrics, across devices are irrelevant and have an F1 of about 70 % and mean Average Precision (mAP) of about 60 %.

Abstract
When compared with traditional local shops where the customer has a personalised service, in large retail departments, the client has to make his purchase decisions independently, mostly supported by the information available in the package. Additionally, people are becoming more aware of the importance of the food ingredients and demanding about the type of products they buy and the information provided in the package, despite it often being hard to interpret. Big shops such as supermarkets have also introduced important challenges for the retailer due to the large number of different products in the store, heterogeneous affluence and the daily needs of item repositioning. In this scenario, the automatic detection and recognition of products on the shelves or off the shelves has gained increased interest as the application of these technologies may improve the shopping experience through self-assisted shopping apps and autonomous shopping, or even benefit stock management with real-time inventory, automatic shelf monitoring and product tracking. These solutions can also have an important impact on customers with visual impairments. Despite recent developments in computer vision, automatic grocery product recognition is still very challenging, with most works focusing on the detection or recognition of a small number of products, often under controlled conditions. This paper discusses the challenges related to this problem and presents a review of proposed methods for retail product label processing, with a special focus on assisted analysis for customer support, including for the visually impaired. Moreover, it details the public datasets used in this topic and identifies their limitations, and discusses future research directions of related fields.

Abstract
The widespread use of smartphones and other low-cost equipment as recording devices, the massive growth in bandwidth, and the ever-growing demand for new applications with enhanced capabilities, made visual data a must in several scenarios, including surveillance, sports, retail, entertainment, and intelligent vehicles. Despite significant advances in analyzing and extracting data from images and video, there is a lack of solutions able to analyze and semantically describe the information in the visual scene so that it can be efficiently used and repurposed. Scientific contributions have focused on individual aspects or addressing specific problems and application areas, and no cross-domain solution is available to implement a complete system that enables information passing between cross-cutting algorithms. This paper analyses the problem from an end-to-end perspective, i.e., from the visual scene analysis to the representation of information in a virtual environment, including how the extracted data can be described and stored. A simple processing pipeline is introduced to set up a structure for discussing challenges and opportunities in different steps of the entire process, allowing to identify current gaps in the literature. The work reviews various technologies specifically from the perspective of their applicability to an end-to-end pipeline for scene analysis and synthesis, along with an extensive analysis of datasets for relevant tasks.