arXiv:2308.10133v2 Announce Type: replace 
Abstract: Face Recognition (FR) technology has made significant strides with the emergence of deep learning. Typically, most existing FR models are built upon Convolutional Neural Networks (CNN) and take RGB face images as the model's input. In this work, we take a closer look at existing FR paradigms from high-efficiency, security, and precision perspectives, and identify the following three problems: (i) CNN frameworks are vulnerable in capturing global facial features and modeling the correlations between local facial features. (ii) Selecting RGB face images as the model's input greatly degrades the model's inference efficiency, increasing the extra computation costs. (iii) In the real-world FR system that operates on RGB face images, the integrity of user privacy may be compromised if hackers successfully penetrate and gain access to the input of this model. To solve these three issues, we propose two novel FR frameworks, i.e., TransFace and TransFace++, which successfully explore the feasibility of applying ViTs and image bytes to FR tasks, respectively. Experiments on popular face benchmarks demonstrate the superiority of our TransFace and TransFace++. Code is available at https://github.com/DanJun6737/TransFace_pp.

يمثل التطور الأخير لـ TransFace++ تقدمًا كبيرًا في تكنولوجيا التعرف على الوجه، مع التركيز على الدقة والكفاءة والأمان. مع استمرار تطور التعلم العميق، تتناول هذه المقاربة الجديدة الثغرات الحرجة في الأطر التقليدية لشبكات CNN، مما يمهد الطريق لتطبيقات أكثر موثوقية وأمانًا في مجالات متنوعة. هذه الابتكار مهم لأنه يعزز موثوقية أنظمة التعرف على الوجه، التي تُستخدم بشكل متزايد في عمليات الأمان والتعريف.

El reciente desarrollo de TransFace++ marca un avance significativo en la tecnología de reconocimiento facial, enfatizando la precisión, la eficiencia y la seguridad. A medida que el aprendizaje profundo continúa evolucionando, este nuevo enfoque aborda vulnerabilidades críticas en los marcos de CNN tradicionales, allanando el camino para aplicaciones más confiables y seguras en diversos campos. Esta innovación es crucial, ya que mejora la fiabilidad de los sistemas de reconocimiento facial, que se utilizan cada vez más en procesos de seguridad e identificación.

Le développement récent de TransFace++ marque une avancée significative dans la technologie de reconnaissance faciale, mettant l'accent sur l'exactitude, l'efficacité et la sécurité. Alors que l'apprentissage profond continue d'évoluer, cette nouvelle approche aborde les vulnérabilités critiques des cadres CNN traditionnels, ouvrant la voie à des applications plus fiables et sécurisées dans divers domaines. Cette innovation est cruciale car elle améliore la fiabilité des systèmes de reconnaissance faciale, de plus en plus utilisés dans les processus de sécurité et d'identification.

The recent development of TransFace++ marks a significant advancement in face recognition technology, emphasizing accuracy, efficiency, and security. As deep learning continues to evolve, this new approach addresses critical vulnerabilities in traditional CNN frameworks, paving the way for more reliable and secure applications in various fields. This innovation is crucial as it enhances the reliability of face recognition systems, which are increasingly used in security and identification processes.

TransFace++: Rethinking the Face Recognition Paradigm with a Focus on Accuracy, Efficiency, and Security

arXiv:2512.02062v1 Announce Type: cross 
Abstract: Deep learning models are used in safety-critical tasks such as automated driving and face recognition. However, small perturbations in the model input can significantly change the predictions. Adversarial attacks are used to identify small perturbations that can lead to misclassifications. More powerful black-box adversarial attacks are required to develop more effective defenses. A promising approach to black-box adversarial attacks is to repeat the process of extracting a specific image area and changing the perturbations added to it. Existing attacks adopt simple rectangles as the areas where perturbations are changed in a single iteration. We propose applying superpixels instead, which achieve a good balance between color variance and compactness. We also propose a new search method, versatile search, and a novel attack method, Superpixel Attack, which applies superpixels and performs versatile search. Superpixel Attack improves attack success rates by an average of 2.10% compared with existing attacks. Most models used in this study are robust against adversarial attacks, and this improvement is significant for black-box adversarial attacks. The code is avilable at https://github.com/oe1307/SuperpixelAttack.git.

تم اقتراح طريقة جديدة تُدعى Superpixel Attack لتعزيز الهجمات العدائية في نماذج التعلم العميق، خاصة في التطبيقات الحرجة للسلامة مثل القيادة الآلية والتعرف على الوجه. تستخدم هذه الطريقة وحدات بكسل فائقة بدلاً من المستطيلات البسيطة لتطبيق الاضطرابات، مما يحسن فعالية الهجمات العدائية والدفاعات.

Se ha propuesto un nuevo método llamado Superpixel Attack para mejorar los ataques adversariales de caja negra en modelos de aprendizaje profundo, especialmente en aplicaciones críticas para la seguridad como la conducción automatizada y el reconocimiento facial. Este enfoque utiliza superpíxeles en lugar de rectángulos simples para aplicar perturbaciones, mejorando la efectividad de los ataques adversariales y las defensas.

Une nouvelle méthode appelée Superpixel Attack a été proposée pour améliorer les attaques adversariales en boîte noire dans les modèles d'apprentissage profond, en particulier dans des applications critiques pour la sécurité comme la conduite automatisée et la reconnaissance faciale. Cette approche utilise des superpixels au lieu de simples rectangles pour appliquer des perturbations, améliorant ainsi l'efficacité des attaques adversariales et des défenses.

A new method called Superpixel Attack has been proposed to enhance black-box adversarial attacks in deep learning models, particularly in safety-critical applications like automated driving and face recognition. This approach utilizes superpixels instead of simple rectangles to apply perturbations, improving the effectiveness of adversarial attacks and defenses.

Superpixel Attack: Enhancing Black-box Adversarial Attack with Image-driven Division Areas

arXiv:2512.02280v1 Announce Type: cross 
Abstract: Artificial intelligence has advanced rapidly across perception, language, reasoning, and multimodal domains. Yet despite these achievements, modern AI systems remain fun- damentally limited in their ability to self-monitor, self-correct, and regulate their behavior autonomously in dynamic contexts. This paper identifies and analyzes seven core deficiencies that constrain contemporary AI models: the absence of intrinsic self- monitoring, lack of meta-cognitive awareness, fixed and non- adaptive learning mechanisms, inability to restructure goals, lack of representational maintenance, insufficient embodied feedback, and the absence of intrinsic agency. Alongside identifying these limitations, we also outline a forward-looking perspective on how AI may evolve beyond them through architectures that mirror neurocognitive principles. We argue that these structural limitations prevent current architectures, including deep learning and transformer-based systems, from achieving robust general- ization, lifelong adaptability, and real-world autonomy. Drawing on a comparative analysis of artificial systems and biological cognition [7], and integrating insights from AI research, cognitive science, and neuroscience, we outline how these capabilities are absent in current models and why scaling alone cannot resolve them. We conclude by advocating for a paradigmatic shift toward cognitively grounded AI (cognitive autonomy) capable of self-directed adaptation, dynamic representation management, and intentional, goal-oriented behavior, paired with reformative oversight mechanisms [8] that ensure autonomous systems remain interpretable, governable, and aligned with human values.

تسلط التطورات الأخيرة في الذكاء الاصطناعي (AI) الضوء على تقدم كبير في القدرات الإدراكية واللغوية والتفكير المتعدد الأبعاد. ومع ذلك، تحدد دراسة جديدة سبع عيوب أساسية في أنظمة الذكاء الاصطناعي الحالية، بما في ذلك عدم وجود مراقبة ذاتية داخلية ووعي ميتا-معرفي، مما يعيق قدرتها على التنظيم الذاتي في البيئات الديناميكية. تشير هذه القيود إلى أن الهياكل الحالية، مثل التعلم العميق وأنظمة التحويل، غير كافية لتحقيق استقلالية معرفية حقيقية.

Los avances recientes en inteligencia artificial (IA) destacan un progreso significativo en las capacidades de percepción, lenguaje, razonamiento y multimodalidad. Sin embargo, un nuevo estudio identifica siete deficiencias fundamentales en los sistemas de IA actuales, incluida la falta de auto-monitoreo intrínseco y conciencia metacognitiva, que obstaculizan su capacidad para autorregularse en entornos dinámicos. Estas limitaciones sugieren que las arquitecturas existentes, como el aprendizaje profundo y los sistemas basados en transformadores, son insuficientes para lograr una verdadera auto…

Les avancées récentes en intelligence artificielle (IA) mettent en lumière des progrès significatifs dans les capacités de perception, de langage, de raisonnement et multimodales. Cependant, une nouvelle étude identifie sept carences fondamentales dans les systèmes IA actuels, notamment l'absence d'auto-surveillance intrinsèque et de conscience métacognitive, qui entravent leur capacité à s'autoréguler dans des environnements dynamiques. Ces limitations suggèrent que les architectures existantes, telles que l'apprentissage profond et les systèmes basés sur des transformateurs, sont insuffisant…

Recent advancements in artificial intelligence (AI) highlight significant progress in perception, language, reasoning, and multimodal capabilities. However, a new study identifies seven core deficiencies in current AI systems, including a lack of intrinsic self-monitoring and meta-cognitive awareness, which hinder their ability to self-regulate in dynamic environments. These limitations suggest that existing architectures, such as deep learning and transformer-based systems, are insufficient for achieving true cognitive autonomy.

Bridging the Gap: Toward Cognitive Autonomy in Artificial Intelligence

arXiv:2512.00125v1 Announce Type: cross 
Abstract: Machine learning, particularly deep learning, is transforming industrial quality inspection. Yet, training robust machine learning models typically requires large volumes of high-quality labeled data, which are expensive, time-consuming, and labor-intensive to obtain in manufacturing. Moreover, defective samples are intrinsically rare, leading to severe class imbalance that degrades model performance. These data constraints hinder the widespread adoption of machine learning-based quality inspection methods in real production environments. Synthetic data generation (SDG) offers a promising solution by enabling the creation of large, balanced, and fully annotated datasets in an efficient, cost-effective, and scalable manner. This paper presents a hybrid SDG framework that integrates simulation-based rendering, domain randomization, and real background compositing to enable zero-shot learning for computer vision-based industrial part inspection without manual annotation. The SDG pipeline generates 12,960 labeled images in one hour by varying part geometry, lighting, and surface properties, and then compositing synthetic parts onto real image backgrounds. A two-stage architecture utilizing a YOLOv8n backbone for object detection and MobileNetV3-small for quality classification is trained exclusively on synthetic data and evaluated on 300 real industrial parts. The proposed approach achieves an mAP@0.5 of 0.995 for detection, 96% classification accuracy, and 90.1% balanced accuracy. Comparative evaluation against few-shot real-data baseline approaches demonstrates significant improvement. The proposed SDG-based approach achieves 90-91% balanced accuracy under severe class imbalance, while the baselines reach only 50% accuracy. These results demonstrate that the proposed method enables annotation-free, scalable, and robust quality inspection for real-world manufacturing applications.

تم تقديم إطار جديد لتوليد البيانات الاصطناعية الهجينة، يجمع بين التصيير القائم على المحاكاة، وتوزيع المجال، وتركيب الخلفيات الحقيقية لتسهيل التعلم بدون إطلاق للنار في فحص الأجزاء المعتمد على الرؤية في البيئات الصناعية. تتناول هذه الطريقة التحديات المتعلقة بالحصول على كميات كبيرة من البيانات المعنونة عالية الجودة، خاصة في السيناريوهات التي تعاني من اختلال شديد في الفئات حيث تكون العينات المعيبة نادرة.

Se ha introducido un nuevo marco de generación de datos sintéticos híbridos que combina renderizado basado en simulación, aleatorización de dominio y composición de fondos reales para facilitar el aprendizaje sin disparos en la inspección de piezas basada en visión en entornos industriales. Este enfoque aborda los desafíos de adquirir grandes volúmenes de datos etiquetados de alta calidad, especialmente en escenarios con un desequilibrio extremo de clases donde las muestras defectuosas son raras.

Un nouveau cadre de génération de données synthétiques hybrides a été introduit, combinant le rendu basé sur la simulation, la randomisation de domaine et le compositing de fonds réels pour faciliter l'apprentissage sans coupure pour l'inspection des pièces basée sur la vision dans des environnements industriels. Cette approche répond aux défis d'acquisition de grands volumes de données étiquetées de haute qualité, en particulier dans des scénarios avec un déséquilibre de classe extrême où les échantillons défectueux sont rares.

A new hybrid synthetic data generation framework has been introduced, combining simulation-based rendering, domain randomization, and real background compositing to facilitate zero-shot learning for vision-based part inspection in industrial settings. This approach addresses the challenges of acquiring large volumes of high-quality labeled data, particularly in scenarios with extreme class imbalance where defective samples are rare.

TransFace++: Rethinking the Face Recognition Paradigm with a Focus on Accuracy, Efficiency, and Security

Was this article worth reading? Share it

VidMax.ai

Lenso.ai

Open Source Surveillance