A Standardized Approach for Skin Detection: Analysis of the Literature and Case Studies

Abstract

Skin detection involves identifying skin and non-skin areas in a digital image and is commonly used in various applications, such as analyzing hand gestures, tracking body parts, and facial recognition. The process of distinguishing between skin and non-skin regions in a digital image is widely used in a variety of applications, ranging from hand-gesture analysis to body-part tracking to facial recognition. Skin detection is a challenging problem that has received a lot of attention from experts and proposals from the research community in the context of intelligent systems, but the lack of common benchmarks and unified testing protocols has hampered fairness among approaches. Comparisons are very difficult. Recently, the success of deep neural networks has had a major impact on the field of image segmentation detection, resulting in various successful models to date. In this work, we survey the most recent research in this field and propose fair comparisons between approaches, using several different datasets. The main contributions of this work are (i) a comprehensive review of the literature on approaches to skin-color detection and a comparison of approaches that may help researchers and practitioners choose the best method for their application; (ii) a comprehensive list of datasets that report ground truth for skin detection; and (iii) a testing protocol for evaluating and comparing different skin-detection approaches. Moreover, we propose an ensemble of convolutional neural networks and transformers that obtains a state-of-the-art performance.

Exploring the Potential of Ensembles of Deep Learning Networks for Image Segmentation

“…Compared with previous work, we standardized the data augmentation step, which was previously implemented in different languages for CNNs (Matlab) and transformers (Python); in this work, we only used the data augmentation created by Matlab. This led to small differences in performance, e.g., the implementation used in this paper of the data augmentation method detailed in [14] obtained an average Dice of 0.891 instead of 0.895, so the method proposed in this work is our suggested ensemble. In addition, this ensemble was tested on four datasets, so we are more confident that the proposed approach will perform well in other datasets.…”

“…• Fusion: the combination of all the nets while varying the DA and LR strategy; • Baseline Ensemble: fusion between nine networks (the same size of Fusion) obtained via retraining DA3-LRc nine times; • SOTAEns: The best ensemble, related to a given topology, previously reported in [13][14][15].…”

“…Additionally, we explore the performance improvement of an ensemble that integrates these methods to evaluate its impact on segmentation performance. This research builds upon prior work [13][14][15], which focused on a limited number of case studies and models. The novelty of our research with respect to previous works lies in the proposal of ensembles that encompass a wide range of network architectures, models, and data augmentation techniques.…”

ICT for disaster-resilient education and training

“…• Marc Baaden, CNRS, France [3] • Liudmyla Bilousova, Academy of Cognitive and Natural Sciences, Ukraine & Israel [4] • Pablo Garcia Bringas, University of Deusto, Spain [5] • Oleksandr Burov, Institute for Digitalisation of Education of the NAES of Ukraine, Ukraine & University of Vienna, Austria [6] • Nadire Cavus, Near East University, North Cyprus [7] • El-Sayed M. El-Horbaty, Ain Shams University, Egypt [8] • Ramón Fabregat, University of Girona, Spain [9] • Irina Georgescu, Bucharest University of Economics, Romania [10] • Mustansar Ali Ghazanfar, University of East London, United Kingdom [11] • Anita Goel, University of Delhi, India [12] • Carina S. Gonzalez, Universidad de La Laguna, Spain [13] • Sven Hartmann, Clausthal University of Technology, Germany [14] • Michail Kalogiannakis, University of Crete, Greece [15] • Arnold Kiv, Ben-Gurion University of the Negev, Israel [16] • Hennadiy Kravtsov, Kherson State University, Ukraine [17] • Olena Kuzminska, National University of Life and Environmental Sciences of Ukraine, Ukraine [18] • Francesco Lelli, Tilburg University, Netherlands [19] • Chung-Sheng Li, PwC, United States [20] • Piotr Lipiński, Technical University of Lodz, Poland [21] • Alessandra Lumini, University of Bologna, Italy [22] • Svitlana Lytvynova, Institute for Digitalisation of Education of the NAES of Ukraine, Ukraine [23] • Maiia Marienko, Institute for Digitalisation of Education of the NAES of Ukraine, Ukraine [24] • Rashid Mehmood, King Abdulaziz University, Saudi Arabia [25] • Iryna Mintii, Institute for Digitalisation of Education of the NAES of Ukraine, Ukraine [26] • Natalia Morze, Borys Grinchenko Kyiv University, Ukraine [27] • Vincenzo Moscato, University of Naples "Federico II", Italia [28] • Thomas Moser, St. Pölten University of Applied Sciences, Austria [29] • Ranesh Kumar Naha, University of Tasmania, Australia [30] • Viacheslav Osadchyi, Borys Grinchenko Kyiv University, Ukraine …”

Comparison of Different Methods for Building Ensembles of Convolutional Neural Networks

“…For instance, deep-learning models have emerged as the state-of-the-art for diagnosing conditions like diabetic retinopathy [11], Alzheimer's disease [12], skin detection [13], gastrointestinal ulcers, and various types of cancer, as demonstrated in recent reviews and studies (see, for instance, [14,15]). Enhancing performance within the medical field carries the greater real impact of this technology compared to other applications.…”