Beyond Fairness in Computer Vision: A Holistic Approach to Mitigating Harms and Fostering Community-Rooted Computer Vision Research

The designers and users of computer vision systems depend on a repertoire of software and hardware building blocks. This article reviews past and future trends in computer vision research and system development. >

Despite many exciting innovations in computer vision, recent studies reveal a number of risks in existing computer vision systems, suggesting results of such systems may be unfair and untrustworthy. Many of these risks can be partly attributed to the use of a training image dataset that exhibits sampling biases and thus does not accurately reflect the real visual world. Being able to detect potential sampling biases in the visual dataset prior to model development is thus essential for mitigating the fairness and trustworthy concerns in computer vision. In this paper, we propose a three-step crowdsourcing workflow to get humans into the loop for facilitating bias discovery in image datasets. Through two sets of evaluation studies, we find that the proposed workflow can effectively organize the crowd to detect sampling biases in both datasets that are artificially created with designed biases and real-world image datasets that are widely used in computer vision research and system development.

This review study looks at two important areas of recent work: improvements in computer vision algorithms and the incorporation of computer graphics into computer vision systems. We explore the development of computer vision in Part 1, highlighting its many uses in various fields as UAV image processing and automobile navigation. Even while real-time processing is essential in many situations, classic local algorithms which are highly valued for their speed often sacrifice image quality in favour of global algorithms. But recent work reveals subtle modifications to matching computations and data gathering techniques that support local algorithms, leading to performance that approaches that of global algorithms with respect to matching rate. In Part 2, an innovative method for combining computer vision and graphics is shown for creating a new telepresence and collaboration platform. Three key features of this system are its smooth integration of the physical and virtual worlds for input and output, its ability to enable remote collaboration between users, and its ability to facilitate interaction between different 3D graphics programs. This system is designed for future high-bandwidth networks and sends real-time fusions of dynamic computer graphics and vision data. This paper highlights the visual dimension and the mutually beneficial relationship between computer graphics and vision for telepresence and collaborative applications. Preliminary studies show promising possibilities for this technology to create immersive spaces suitable for various cooperative tasks across high-bandwidth networks. Our goal in doing this thorough assessment is to clarify the course of these

Despite significant efforts to democratize artificial intelligence (AI), computer vision which is a sub-field of AI, still lags in Africa. A significant factor to this, is the limited access to computing resources, datasets, and collaborations. As a result, Africa’s contribution to top-tier publications in this field has only been 0.06% over the past decade. Towards improving the computer vision field and making it more accessible and inclusive, this study analyzes 63,000 Scopus-indexed computer vision publications from Africa. We utilize large language models to automatically parse their abstracts, to identify and categorize topics and datasets. This resulted in listing more than 100 African datasets. Our objective is to provide a comprehensive taxonomy of dataset categories to facilitate better understanding and utilization of these resources. We also analyze collaboration trends of researchers within and outside the continent. Additionally, we conduct a large-scale questionnaire among African computer vision researchers to identify the structural barriers they believe require urgent attention. In conclusion, our study offers a comprehensive overview of the current state of computer vision research in Africa, to empower marginalized communities to participate in the design and development of computer vision systems.

The book, Deep Learning for Computer Vision Systems: Algorithms and Applications, is book for new era where we have tried to collect a variety of studies that actually brought a rising effect of deep learning on computer vision and intelligent systems. This book aims to represents and introduces basic concepts and real-world applications of deep learning. The chapters are organised in a way that make the readers to understand basic ideas to work with deep learning to advanced level topics. The book comprises of introductory explanation on deep learning and computer vision topics that set the momentum for understand contemporary algorithms, while later chapters cover the applications that help to understand how it can be applied in practice. Few chapters focus on application areas such as medical image analysis for lung and breast cancer detection, underwater object detection, and object localization, emphasizing how deep learning models can bring decision-making. In addition, the book includes vital ethical reflections and safety challenges related to computer vision systems. This book is proposed to assist students, researchers, academicians, and industry professionals who seek out both academic acumens and practical knowledge in the field of deep learning and computer vision. We are hopeful that this book will support learning, encourage research that stimulate the development of innovative technologies.

Manual analysis of rice and corn is done by visually inspecting each grain and classifying according to their respective categories. This method is subjective and tedious leading to errors in analysis. Computer vision could be used to analyze quality of rice and corn by developing models that correlate shape and color features with various classification. The PhilMech low-cost computer vision system (CVS) was developed to analyze the quality of rice and corn. It is composed of an ordinary scanner as the image acquisition device and a computer with image-processing software. The performance of the CVS was compared to the traditional manual method being adopted by the National Food Authority (NFA) and the Agricultural Machinery Testing and Evaluation Center (AMTEC). The performance testing and evaluation showed that the accuracy of obtaining the results in classifying rice and corn using the CVS was comparable to the manual method of analysis. But, the processing time to complete the analysis using the CVS technology (6-7 minutes) was 5-8 times faster compared to the manual method (30-60 minutes). The developed CVS will automate the existing practice in determining the milling quality of brown rice, milled rice and yellow corn and minimize the tedious and subjective manual method of evaluation.

Image processing techniques are increasingly applied in sorting applications of agricultural products. This work has assessed the use of image processing for inspecting surface color of two Thai mango cultivars. A computer vision system (CVS) was developed and experiments were conducted to monitor peel color change during the ripening process. Conversion of RGB to CIE-LAB values was done via image processing and prediction models were developed to estimate color parameters from CVS data. Performance evaluations showed insufficient prediction for L values (R2 = 0.42-0.58), but better results for A and B values (R2 = 0.90-0.95 and 0.80-0.82, respectively). Compared to the calculated color values hue angle and chroma, a yellowness index computed from intermediate XYZ values was found to be much more adept at accurately predicting peel color from CVS data. Correlations were strong for both cultivars (R2 = 0.93 for ‘Nam Dokmai’ and R2 = 0.95 for ‘Maha Chanok’). Results from classification analysis indicated satisfactory results for classifying fruits according to ripeness based on yellowness. Success rates of true positives in the categories unripe, ripe and overripe ranged 72%-92% for ‘Nam Dokmai’ and 98%-100% for ‘Maha Chanok’. Therefore, it was shown that the CVS was capable of producing accurate color values for the two mango cultivars investigated. The findings of this study can be incorporated for development of a robust system for quality prediction and establishment of a CVS for automatic grading and sorting of mangos. Keywords: mango, peel color, computer vision, image processing, fruit quality, Thailand DOI: 10.3965/j.ijabe.20160901.1861 Citation: Nagle M, Intani K, Romano G, Mahayothee B, Sardsud V, Muller J. Determination of surface color of ‘all yellow’ mango cultivars using computer vision. Int J Agric & Biol Eng, 2016; 9(1): 42－50.

Recent developments in computer vision and conversational systems have provided the AI community with novel perspectives towards improving the cognitive capabilities of engaging socially assistive robots. We show how to develop conversational skills for a hospital receptionist robot that incorporates social conversation based on visual information as well as task-based dialog. Fusing the traditional modular conversational system architecture with recent developments in computer vision and scene graph research, our agent (called ‘ViCA’) supports both visual question answering and social conversational capabilities based on the visual scene. In particular, our agent can provide guidance to users by locating visible objects in the room and can engage in social dialog using visual prompts, such as the user’s clothing or possessions. We conduct a comprehensive online evaluation study with 21 participants, showcasing that the ViCA system is perceived as both helpful and entertaining.

Research and development are in progress in signal processing for explosive detection for airport security. The recent advancements in real-time computer vision, however, will provide possibilities of far more new and innovative ways to augment airport's human security force. For example, a computer vision system with real-time object tracking capability can monitor the behavior of large groups of people, computers can keep detailed records of the flow of people through certain areas, allowing security personnel to predict and quickly respond to crowd or isolation situations, as well as identifying potential suspects through suspicious movement patterns. Further, such a system would enable airport security officers to accurately monitor a particular suspect's movements within the airport, relayed from camera to camera, without alerting him to their observation or diverting manpower from other areas. Computerized face recognition and facial expression extraction are other actively researched areas in computer vision that have potential for airport security applications. I present an overview of the state-of-the-art computer vision research than can have relevance to airport security, and discuss how they can be applied and what aspects of that technology need to be researched further before they can be put into practical application.© (1993) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Animal detection and classifications using computer vision systems have been practically unusable in uncontrolled contexts for years, due to concerns about the accuracy of the algorithms used. As a result, object detection, recognition, and classification, among other things, are receiving a lot of attention. Visual monitoring of animals in their natural habitat is currently one of the most popular methods in the field of computer vision (CV) systems. However, real-time animal detection and classification methods are still unavailable. Deep Learning improvements in Computer Vision have been built and developed through time, primarily over one specific method - a Convolutional Neural Network(CNN). This project is proposed using CNN to detect and classify animals in digital images. The feature of the animal in the input image is extracted to aid decision making, which is then plugged into the classification part for model analysis. According to the experimental results, the proposed algorithm has a positive effect on overall animal classification performance. An accuracy of around 64% was achieved when the image size was 50 × 50..

Ignorance, myopia, and naiveté in computer vision systems

A computer vision and robotic system for an intelligent workstation

Background. Chest low-dose computed tomography (LDCT) is used in lung cancer screening, but the study data can also be used to assess the liver condition, including the hepatic steatosis (HS) detection. However, radiologists often do not pay attention to liver changes due to the focus on the chest. Objective: to determine the prevalence of HS during chest LDST among lung cancer screening patients using a computer vision (CV) system. Material and methods. For a retrospective study, 300 chest LDCT were taken from Moscow lung cancer screening in 2018–2020. Hepatic attenuation analysis was performed by CV, the values < 40 HU were considered as HS. The text protocols of CT scans were analysed and compared with decreased hepatic attenuation revealed by CV system. Results. 291 patients were analysed, the median age for the sample was 65 [61; 70] years. The mean hepatic attenuation was 55.6 ± 14.8 HU. Hepatic attenuation < 40 HU was found in 13% patients (23 (16.1%) males and 14 (9.5%) females), a statistically significant difference was revealed among these patients (p = 0.04). Six (4.2%) males and 4 (2.7%) females were at risk for HS (40–45 HU). The examination of text protocols showed no pathology discovered in all cases. Conclusion. The prevalence of CT signs for HS among the lung cancer screening group in Moscow was 13%. The absence of HS in text protocols highlights the importance of using CV systems in the routine practice.

The role of perceptual organization in computer vision systems is explored. This is done from four vantage points. A brief history of perceptual organization research in both humans and computer vision is offered. A classificatory structure in which to cast perceptual organization research to clarify both the nomenclature and the relationships among the many contributions is proposed. The perceptual organization work in computer vision in the context of this classificatory structure is reviewed. The array of computational techniques applied to perceptual organization problems in computer vision is surveyed.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

A review of deep learning algorithms for computer vision systems in livestock