Deep-learning-based image inpainting technology has achieved remarkable visual consistency but is vulnerable to malicious use. Existing detection methods overlook semantic inconsistencies between targets and backgrounds, leading to ambiguous results due to low discriminability. To tackle these challenges, we draw inspiration from human strategies in visual tasks, which involve initially assigning uncertainty across the entire input and subsequently concentrating on highly uncertain regions using prior knowledge like boundary information. Building on this, we propose a Dual Information Guided Network (DIGNet). It combines object-background semantic modulation with uncertainty to precisely locate inpainting regions. This is the first work to address inpainting prediction inaccuracies by considering both edge uncertainty and semantic inconsistency. DIGNet consists of three key parts: the Edge Uncertainty Awareness Module (EUAM), the Edge Correction Module (ECM) based on semantic differences, and the Dual Information Guided Interaction Module (DIGIM). We use semantic inconsistency to get edge constraints and quantify uncertainty as feature variance to guide mainstream feature maps. The DIGIM effectively fuses guide information for accurate predictions. Comprehensive experiments show that our method outperforms existing CNN-based approaches. Specifically, it improves the F1 Score by at least 0.31% and the IOU by at least 0.19% on multiple datasets.

Abstract Although the work of identifying animal and plant objects with highly similar patterns (e.g., texture, intensity, colour, etc.) to the background has recently attracted more research interest but rarely involves the military complex environment. Design an efficient camouflage small‐scale object detection algorithm that is capable of quickly discriminating the objects in open scenes from a long distance, to pre‐empt the enemy. In this work, we first recognize the fact that existing open‐source training datasets are scarce, and we have created a specific disguised people benchmark covering multiple scenes and weather conditions. Second, because the severe corruption of camouflage capabilities and chaotic scenes in the open battlefield on detailed features and generalization intensifies the challenge of feature extraction from a long‐range perspective, we propose a novel end‐to‐end Small‐scale open scene Camouflage Object Detection Network, called SM‐CODN. Inspired by the characteristics of biological brain partition, a multi‐domain partition module (MPM) with domain‐decoupling is proposed to enable specific knowledge learning for samples with obvious discrepancies in camouflage domain distribution. Concurrent with our work, we have designed a multi‐scale fusion module (MFM) to strengthen the semantic features related to small‐scale disguised objects. Moreover, due to the convergence direction of the detector in reasoning being inconsistent, a feature separation enhancement module (FSEM) is also proposed. Experimental results show that SM‐CODN surpasses many classic object detection methods and shows strong competitiveness compared with state‐of‐the‐art ones.

Abstract In recent years, convolutional neural networks are playing an increasingly important role in the field of object detection. However, the complex background of the detected image, the limited receptive field by the fixed geometry of the convolution kernel when building the model, and the positioning and pooling deviation from the region of interest are still important factors that affect the detection accuracy. In this paper, an improved algorithm is proposed for target detection based on Faster regional convolutional neural network. In the bounding box positioning phase, an improved interpolation algorithm‐Newton's parabolic interpolation is proposed instead of bilinear interpolation, after ROI size normalized by extending a parallel branch of tensor to weaken the negative impact of complex background on prospects in the phase of feature extraction using neural network recently popular attention CBAM mechanism model and the deformable convolution. Without bells and whistles, a series of experiments show that our method has higher target detection accuracy on the datasets PASCAL VOC2007, VOC2012, COCO 2014 and DIOR. Hence, the method is effective for actual target recognition tasks in complex background environments. The authors hope that the method will contribute to future research. Code has been made available at: https://github.com/liumaozhen‐lmz/Faster_R‐CNN_Attention.git.