Adult content images have a detrimental effect on Internet users, a significant number of whom are minors. Therefore, it is essential to control and detecting adult content images using multimedia processing and computer vision techniques. Previous studies have typically focused on manual-engineered visual features that may be difficult to detect and analyze. This paper presents a new model that employs deep convolutional neural networks within a Gaussian-Bernoulli limited-time, for adult content image recognition of a wide variety in a precise and effective manner. There are various layers within Convolutional Neural Networks for feature extraction and classification. Gaussian-Bernoulli limited-time was used for feature extraction to describe the images, and these features were summarized using the Boltzmann machine limited in the feature summary phase. The benefit of such an approach is convenience in carrying out feature extraction. Additionally, when tested on the most modern criterion dataset, this finding is believed to be more precise compared to other state-of-the-art approaches. The results obtained prove that the proposed approach leads to a higher efficiency.

Face recognition systems typically face actual challenges like facial pose, illumination, occlusion, and ageing that significantly impact the recognition accuracy. In this paper, a robust face recognition system that uses Multi-task Cascaded Convolutional Networks (MTCNN) for face detection and face alignment with an enhanced FaceNet for facial embedding extraction is presented. The enhanced FaceNet uses attention mechanisms to achieve more discriminative facial embeddings, especially in challenging scenarios. In addition, an Adaptive Feature Fusion module synthetically combines identity-specific embeddings with context information such as pose, lighting, and presence of masks, hence enhancing robustness and accuracy. Training takes place using the CelebA dataset, and the test is conducted independently on LFW and IJB-C to enable subject-disjoint evaluation. CelebA has over 200,000 faces of 10,177 individuals, LFW consists of 13,000+ faces of 5749 individuals in unconstrained conditions, and IJB-C has 31,000 faces and 117,000 video frames with extreme pose and occlusion changes. The system introduced here achieves 99.6% on CelebA, 94.2% on LFW, and 91.5% on IJB-C and outperforms baselines such as simple MTCNN-FaceNet, AFF-Net, and state-of-the-art models such as ArcFace, CosFace, and AdaCos. These findings demonstrate that the proposed framework generalizes effectively between datasets and is resilient in real-world scenarios.

This paper proposes a new model that improves face recognition with challenges such as illumination variation and noise in facial images. The model combines state-of-the-art illumination normalization by Retinex theory with median filtering for improved noise reduction and accuracy of recognition. Using pre-trained VGG-16 weights, the model applies parts of the Boltzmann machine to extract intricate features from faces while minimizing overfitting. Further, this paper combines zero-shot facial expression recognition methodologies and illumination-robust face recognition within one facial analysis framework. The combination of models for pre-processing with large language models and more advanced feature extraction allows the model to perform well on various benchmarks from FERET to LFW, CelebA, and VGGFace2 datasets. These results have identified that the integration of superior normalization techniques along with zero-shot learning enhances face recognition and expression analysis in difficult, practical conditions. This solution is feasible for a practical situation and robust against real-world variations in illumination and noise. Future work will enhance computational efficiency and expand the applicability to various datasets.