Adversarial attacks that mislead deep neural networks (DNNs) into making incorrect predictions can also be implemented in the physical world. However, most of the existing adversarial camouflage textures that attack object detection models only consider the effectiveness of the attack, ignoring the stealthiness of adversarial attacks, resulting in the generated adversarial camouflage textures appearing abrupt to human observers. To address this issue, we propose a style transfer module added to an adversarial texture generation framework. By calculating the style loss between the texture and the specified style image, the adversarial texture generated by the model is guided to have good stealthiness and is not easily detected by DNNs and human observers in specific scenes. Experiments have shown that in both the digital and physical worlds, the vehicle full coverage adversarial camouflage texture we create has good stealthiness and can effectively fool advanced DNN object detectors while evading human observers in specific scenes.

Infrared images have garnered significant interest due to their superior performance, driving extensive research on visible-to-infrared image translation. However, existing cross-domain generation methods lack specialization for infrared image generation under aerial perspective, leading to distribution inconsistencies between synthetic and real infrared images and failing to mitigate challenges like small-target blurring and background interference under aerial views. To address these issues, we propose an RGB-to-infrared image generation method based on the Brownian bridge diffusion model for aerial perspective. Technically, we optimize the diffusion coefficient and variance scheduling of the Brownian bridge by introducing a parabolic function, design a Laplacian of Gaussian (LOG) loss that fuses high-, medium-, and low-frequency features, and construct two core modules: a modality enhancement module that integrates spectral involution and cross-modal fusion, and an information guidance module based on wavelet decomposition. Experimental results demonstrate state-of-the-art performance: the method achieves a PSNR of 15.06 and an SSIM of 49.47, which are 1.5% and 1.2% higher than the suboptimal baseline BBDM-VQ4, respectively; its FID is reduced to 36.83, representing a 25.6% decrease compared to BBDM-VQ4, and its LPIPS is 2.0% lower than that of BBDM-VQ4. This approach effectively eliminates distribution biases induced by small-target blurring and background interference under aerial perspective while ensuring the semantic consistency of generated infrared images.

Camouflaged object detection (COD) is a challenging task that involves identifying objects that closely resemble their background. In order to detect camouflaged objects more accurately, we propose a diffusion model for the COD network called DMNet. DMNet formulates COD as a denoising diffusion process from noisy boxes to prediction boxes. During the training stage, random boxes diffuse from ground-truth boxes, and DMNet learns to reverse this process. In the sampling stage, DMNet progressively refines random boxes to prediction boxes. In addition, due to the camouflaged object’s blurred appearance and the low contrast between it and the background, the feature extraction stage of the network is challenging. Firstly, we proposed a parallel fusion module (PFM) to enhance the information extracted from the backbone. Then, we designed a progressive feature pyramid network (PFPN) for feature fusion, in which the upsample adaptive spatial fusion module (UAF) balances the different feature information by assigning weights to different layers. Finally, a location refinement module (LRM) is constructed to make DMNet pay attention to the boundary details. We compared DMNet with other classical object-detection models on the COD10K dataset. Experimental results indicated that DMNet outperformed others, achieving optimal effects across six evaluation metrics and significantly enhancing detection accuracy.

The task of camouflaged object segmentation (COS) is a challenging endeavor that entails the identification of objects that closely blend in with their surrounding background. Furthermore, the camouflaged object’s obscure form and its subtle differentiation from the background present significant challenges during the feature extraction phase of the network. In order to extract more comprehensive information, thereby improving the accuracy of COS, we propose a diffusion model for a COS network that utilizes frequency domain information as auxiliary input, and we name it FreDiff. Firstly, we proposed a frequency auxiliary module (FAM) to extract frequency domain features. Then, we designed a Global Fusion Module (GFM) to make FreDiff pay attention to the global features. Finally, we proposed an Upsample Enhancement Module (UEM) to enhance the detailed information of the features and perform upsampling before inputting them into the diffusion model. Additionally, taking into account the specific characteristics of COS, we develop the specialized training strategy for FreDiff. We compared FreDiff with 17 COS models on the four challenging COS datasets. Experimental results showed that FreDiff outperforms or is consistent with other state-of-the-art methods under five evaluation metrics.

The robustness of vision-language agents in real-world environments depends critically on the reliability of their underlying object detectors. Adversarial camouflage has emerged as a promising approach for executing multi-view attacks against these detectors, yet its effectiveness on full-scale, complex real-world objects remains largely unverified. Existing physical validations are predominantly limited to scaled models, leaving a significant gap in understanding real-world threats. Building upon prior digital simulations and scaled-model experiments, this study presents the first systematic quantitative evaluation of full-coverage adversarial camouflage applied to an actual vehicle. We transfer textures generated in the digital domain to a real vehicle and conduct extensive outdoor tests under varying lighting conditions and viewing angles, including aerial perspectives. The attack performance is benchmarked against multiple mainstream detectors. Our results reveal a discrepancy between digital and physical effectiveness. While the camouflage exhibits a measurable attack capability in the physical world, its impact is significantly attenuated by factors including texture transfer loss, environmental interference, and detector robustness. By providing empirical data and a detailed analysis of these limiting factors, this work offers actionable insights for designing more resilient vision-language perception systems against physical-world adversarial threats.