arXiv:2510.25175v1 Announce Type: new 
Abstract: In recent years, test-time adaptive object detection has attracted increasing attention due to its unique advantages in online domain adaptation, which aligns more closely with real-world application scenarios. However, existing approaches heavily rely on source-derived statistical characteristics while making the strong assumption that the source and target domains share an identical category space. In this paper, we propose the first foundation model-powered test-time adaptive object detection method that eliminates the need for source data entirely and overcomes traditional closed-set limitations. Specifically, we design a Multi-modal Prompt-based Mean-Teacher framework for vision-language detector-driven test-time adaptation, which incorporates text and visual prompt tuning to adapt both language and vision representation spaces on the test data in a parameter-efficient manner. Correspondingly, we propose a Test-time Warm-start strategy tailored for the visual prompts to effectively preserve the representation capability of the vision branch. Furthermore, to guarantee high-quality pseudo-labels in every test batch, we maintain an Instance Dynamic Memory (IDM) module that stores high-quality pseudo-labels from previous test samples, and propose two novel strategies-Memory Enhancement and Memory Hallucination-to leverage IDM's high-quality instances for enhancing original predictions and hallucinating images without available pseudo-labels, respectively. Extensive experiments on cross-corruption and cross-dataset benchmarks demonstrate that our method consistently outperforms previous state-of-the-art methods, and can adapt to arbitrary cross-domain and cross-category target data. Code is available at https://github.com/gaoyingjay/ttaod_foundation.

تقدم ورقة جديدة على arXiv نهجًا مبتكرًا للكشف عن الكائنات القابلة للتكيف في وقت الاختبار، والذي يكتسب اهتمامًا متزايدًا بسبب قدرته على التكيف مع سيناريوهات العالم الحقيقي. تتناول هذه الطريقة القيود المفروضة على التقنيات الحالية التي تفترض أن مجالات المصدر والهدف متطابقة. من خلال اقتراح نموذج أساسي، يهدف المؤلفون إلى تحسين قابلية تكيف أنظمة الكشف عن الكائنات، مما يجعلها أكثر فعالية في بيئات متنوعة. هذه الخطوة مهمة لأنها قد تؤدي إلى تحسين الأداء في تطبيقات مختلفة، من المركبات المستقلة إلى أنظمة المراقبة.

Un nuevo artículo en arXiv presenta un enfoque innovador para la detección de objetos adaptativa en tiempo de prueba, que está ganando atención por su capacidad para adaptarse a escenarios del mundo real. Este método aborda las limitaciones de las técnicas existentes que suponen que los dominios de origen y destino son idénticos. Al proponer un modelo fundamental, los autores buscan mejorar la adaptabilidad de los sistemas de detección de objetos, haciéndolos más efectivos en diversos entornos. Este avance es significativo, ya que podría llevar a un mejor rendimiento en diversas aplicaciones, desde vehículos autónomos hasta sistemas de vigilancia.

Un nouvel article sur arXiv présente une approche révolutionnaire de la détection d'objets adaptative en temps de test, qui suscite un intérêt croissant pour sa capacité à s'adapter à des scénarios réels. Cette méthode répond aux limites des techniques existantes qui supposent que les domaines source et cible sont identiques. En proposant un modèle fondamental, les auteurs visent à améliorer l'adaptabilité des systèmes de détection d'objets, les rendant plus efficaces dans des environnements divers. Cette avancée est significative car elle pourrait conduire à de meilleures performances dans diverses applications, des véhicules autonomes aux systèmes de surveillance.

A new paper on arXiv introduces a groundbreaking approach to test-time adaptive object detection, which is gaining traction for its ability to adapt to real-world scenarios. This method addresses the limitations of existing techniques that assume source and target domains are identical. By proposing a foundation model, the authors aim to enhance the adaptability of object detection systems, making them more effective in diverse environments. This advancement is significant as it could lead to improved performance in various applications, from autonomous vehicles to surveillance systems.

Test-Time Adaptive Object Detection with Foundation Model

arXiv:2511.19117v1 Announce Type: new 
Abstract: The miniaturization of thermal sensors for mobile platforms inherently limits their spatial resolution and textural fidelity, leading to blurry and less informative images. Existing thermal super-resolution (SR) methods can be grouped into single-image and RGB-guided approaches: the former struggles to recover fine structures from limited information, while the latter relies on accurate and laborious cross-camera calibration, which hinders practical deployment and robustness. Here, we propose 3M-TI, a calibration-free Multi-camera cross-Modality diffusion framework for Mobile Thermal Imaging. At its core, 3M-TI integrates a cross-modal self-attention module (CSM) into the diffusion UNet, replacing the original self-attention layers to adaptively align thermal and RGB features throughout the denoising process, without requiring explicit camera calibration. This design enables the diffusion network to leverage its generative prior to enhance spatial resolution, structural fidelity, and texture detail in the super-resolved thermal images. Extensive evaluations on real-world mobile thermal cameras and public benchmarks validate our superior performance, achieving state-of-the-art results in both visual quality and quantitative metrics. More importantly, the thermal images enhanced by 3M-TI lead to substantial gains in critical downstream tasks like object detection and segmentation, underscoring its practical value for robust mobile thermal perception systems. More materials: https://github.com/work-submit/3MTI.

إن إدخال 3M-TI، وهو إطار عمل متعدد الكاميرات بدون معايرة، يمثل تقدمًا كبيرًا في تكنولوجيا التصوير الحراري المحمول. تتضمن هذه الطريقة المبتكرة دمج وحدة انتباه متعددة الأبعاد في UNet الخاص بالانتشار، مما يسمح بمحاذاة الميزات الحرارية وRGB دون الحاجة إلى معايرة معقدة، مما يحسن جودة الصورة وتفاصيلها.

La introducción de 3M-TI, un marco de difusión multi-cámara sin calibración, marca un avance significativo en la tecnología de imágenes térmicas móviles. Este enfoque innovador integra un módulo de atención cruzada en el UNet de difusión, permitiendo la alineación de características térmicas y RGB sin necesidad de calibración compleja de cámaras, mejorando así la calidad y el detalle de las imágenes.

L'introduction de 3M-TI, un cadre de diffusion multi-caméras sans calibration, marque une avancée significative dans la technologie d'imagerie thermique mobile. Cette approche innovante intègre un module d'attention croisée dans le diffusion UNet, permettant l'alignement des caractéristiques thermiques et RGB sans nécessiter de calibration complexe des caméras, améliorant ainsi la qualité et le détail des images.

The introduction of 3M-TI, a calibration-free Multi-camera cross-Modality diffusion framework, marks a significant advancement in mobile thermal imaging technology. This innovative approach integrates a cross-modal self-attention module into the diffusion UNet, enabling the alignment of thermal and RGB features without the need for complex camera calibration, thus enhancing image quality and detail.

Test-Time Adaptive Object Detection with Foundation Model

Was this article worth reading? Share it

Open Source Surveillance

Dyad

Cometapi-e0d0fd