Generative AI in Depth: A Survey of Recent Advances, Model Variants, and Real-World Applications

arXiv — cs.CV•Tuesday, October 28, 2025 at 4:00:00 AM

Recent advancements in generative AI, particularly through models like GANs, VAEs, and DMs, are transforming how we create high-quality content in fields such as image and video synthesis. This surge in capability not only showcases the power of deep learning but also highlights the growing public interest and adoption of these technologies. As these models evolve, they promise to unlock even more innovative applications, making this an exciting time for both creators and consumers.

— via World Pulse Now AI Editorial System

Read Original

Was this article worth reading? Share it

Magicley AI

Access a suite of AI generators for all your creative and productivity tasks.

AI & DataView app details

Videotok

Generate viral videos automatically using advanced AI technology.

AI & DataView app details

AiReelGenerator.com

Generate and publish faceless videos automatically with AI.

AI & DataView app details

KissGen AI

Generate AI videos and images with advanced tools for creative projects.

Creative & DesignView app details

RandomGenerator.AI

Generate random data, images, and text instantly with AI for creative projects and decisions.

Business & ProductivityView app details

Bulk Image Generation AI

Generate over 100 professional-grade images in just 20 seconds with AI.

AI & DataView app details

Continue Readings

arXiv — stat.ML2 days ago

If generative AI is the answer, what is the question?

NeutralArtificial Intelligence

Generative AI has evolved from generating text and images to encompassing audio, video, computer code, and molecular structures. This expansion raises critical questions about the nature of generative AI as a distinct machine learning task, linking it to prediction, compression, and decision-making processes. The article surveys five major generative model families, including autoregressive models and diffusion models, and discusses the implications of these technologies.

Read full article

via arXiv — stat.ML

arXiv — cs.LG2 days ago

CIEGAD: Cluster-Conditioned Interpolative and Extrapolative Framework for Geometry-Aware and Domain-Aligned Data Augmentation

PositiveArtificial Intelligence

The proposed CIEGAD framework aims to enhance data augmentation in deep learning by addressing the challenges of data scarcity and label imbalance, which often lead to misclassification and unstable model behavior. By employing cluster conditioning and hierarchical frequency allocation, CIEGAD systematically improves both in-distribution and out-of-distribution data regions.

Read full article

via arXiv — cs.LG

arXiv — cs.LG2 days ago

Metacognitive Sensitivity for Test-Time Dynamic Model Selection

PositiveArtificial Intelligence

A new framework for evaluating AI metacognition has been proposed, focusing on metacognitive sensitivity, which assesses how reliably a model's confidence predicts its accuracy. This framework introduces a dynamic sensitivity score that informs a bandit-based arbiter for test-time model selection, enhancing the decision-making process in deep learning models such as CNNs and VLMs.

Read full article

via arXiv — cs.LG

arXiv — cs.LG2 days ago

PMB-NN: Physiology-Centred Hybrid AI for Personalized Hemodynamic Monitoring from Photoplethysmography

PositiveArtificial Intelligence

A new study introduces the Physiological Model-Based Neural Network (PMB-NN), a hybrid AI approach designed for personalized hemodynamic monitoring using photoplethysmography (PPG). This method integrates deep learning with a Windkessel model to enhance blood pressure estimation and improve interpretability, addressing limitations in existing data-driven techniques.

Read full article

via arXiv — cs.LG

arXiv — cs.LG2 days ago

Symmetry in Neural Network Parameter Spaces

NeutralArtificial Intelligence

A recent survey published on arXiv explores the concept of symmetry in neural network parameter spaces, highlighting how modern deep learning models exhibit significant overparameterization. This redundancy is largely attributed to symmetries that maintain the network's output unchanged, influencing optimization and learning dynamics.

Read full article

via arXiv — cs.LG

arXiv — cs.CV2 days ago

Robust Multi-Disease Retinal Classification via Xception-Based Transfer Learning and W-Net Vessel Segmentation

PositiveArtificial Intelligence

A recent study has introduced a robust multi-disease retinal classification system utilizing Xception-based transfer learning and W-Net vessel segmentation, addressing the increasing incidence of vision-threatening ocular conditions. This approach combines deep feature extraction with interpretable image processing to enhance the accuracy of automated diagnoses.

Read full article

via arXiv — cs.CV

arXiv — cs.CV2 days ago

AlcheMinT: Fine-grained Temporal Control for Multi-Reference Consistent Video Generation

PositiveArtificial Intelligence

AlcheMinT has been introduced as a unified framework for subject-driven video generation, enhancing fine-grained temporal control over subject appearance and disappearance through explicit timestamp conditioning. This advancement addresses limitations in existing methods, making it suitable for applications like compositional video synthesis and controllable animation.

Read full article

via arXiv — cs.CV

arXiv — cs.CV2 days ago

SpotLight: Shadow-Guided Object Relighting via Diffusion

PositiveArtificial Intelligence

The recent introduction of SpotLight, a method for shadow-guided object relighting via diffusion models, allows for precise control over lighting in neural rendering without additional training. By injecting a coarse shadow hint, the method enables accurate shading of virtual objects in images, harmonizing them with their backgrounds.

Read full article

via arXiv — cs.CV

Ready to build your own newsroom?

Subscribe to unlock a personalised feed, podcasts, newsletters, and notifications tailored to the topics you actually care about