Skip to main content
GameDev.net gamedev.net

Deep Learning News

The latest Deep Learning coverage curated for game developers.

The History of AI Cuts Through Visual Computing

AI in games didn’t start with chatbots; it grew out of decades of visual computing, from rule-based game logic to digital humans, ImageNet, and deep learning. That lineage matters because …

Blog SIGGRAPH · 2 months ago
High-Fidelity 3D Geometric Reconstruction of Pelvic Organs from MRI: A Hybrid Deep Learning and Iterative Optimization Approach

A hybrid deep-learning plus iterative-optimization pipeline reconstructs patient-specific pelvic organ meshes from MRI with better geometric fidelity than prior ML-only approaches. The paper reports lower Chamfer Distance, higher Dice scores, …

Research arXiv cs.GR · 3 months ago
Applying Deep Learning for cockpit segmentation in the context of mixed reality

A mixed-reality cockpit segmentation paper reports roughly 90% accuracy using U-Net and DeepLabV3+ on images from a CAT793F off-highway truck simulator. For teams building MR training or vehicle sims, the …

Research arXiv cs.GR · 3 months, 1 week ago
Image-aware Layout Generation with User Constraints for Poster Design

A new deep-learning model for automatic layout generation in poster design allows for user-defined constraints, enhancing creative workflows for designers. This innovation could significantly streamline the design process, ensuring layouts …

Research arXiv cs.GR · 4 months ago
Animator-Centric Skeleton Generation on Objects with Fine-Grained Details

A new animator-centric skeleton generation framework promises to enhance 3D asset animation by addressing the limitations of current deep learning methods. With a dataset of over 82,000 rigged meshes, this …

Research arXiv cs.GR · 4 months, 4 weeks ago
FatigueFusion: Latent Space Fusion for Fatigue-Driven Motion Synthesis

FatigueFusion introduces a novel deep-learning architecture that allows developers to synthesize fatigue-driven motion in 3D animations. This tool offers a significant advancement for graphics programmers and animators, enabling the creation …

Research arXiv cs.GR · 5 months ago
Cross-Scenario Deraining Adaptation with Unpaired Data: Superpixel Structural Priors and Multi-Stage Pseudo-Rain Synthesis

A new framework for image deraining adaptation is set to enhance outdoor surveillance and autonomous driving systems. By using unpaired data, this method achieves significant performance improvements, with PSNR gains …

Research arXiv cs.GR · 5 months, 3 weeks ago
SDGraph: Multi-Level Sketch Representation Learning by Sparse-Dense Graph Architecture

The introduction of SDGraph marks a significant advancement in sketch representation learning, particularly for artists and designers. By improving accuracy in classification and retrieval tasks by up to 2.30%, this …

Research arXiv cs.GR · 6 months, 1 week ago
Deep Accurate Solver for the Geodesic Problem

A new deep learning method enhances geodesic distance calculations on surfaces, achieving third-order accuracy. This advancement is particularly relevant for graphics programmers who rely on precise distance metrics for rendering …

Research arXiv cs.GR · 6 months, 3 weeks ago
Sketch2Scene: Automatic Generation of Interactive 3D Game Scenes from User's Casual Sketches

A new deep-learning approach allows developers to create interactive 3D game scenes from simple sketches. This innovation streamlines the content creation process, making it easier for designers and artists to …

Research arXiv cs.GR · 7 months, 1 week ago
MACS: Multi-source Audio-to-image Generation with Contextual Significance and Semantic Alignment

The introduction of MACS marks a significant advancement in audio-to-image generation, allowing for multi-source audio inputs to create richer visual content. This innovation is particularly relevant for audio engineers and …

Research arXiv cs.GR · 9 months, 1 week ago
Rethinking Few-Shot Image Fusion: Granular Ball Priors Enable General-Purpose Deep Fusion

The introduction of Granular Ball Priors in image fusion could significantly streamline workflows for graphics programmers and artists. By allowing neural networks to learn fusion rules with minimal data, this …

Research arXiv cs.GR · 9 months, 1 week ago
ENTIRE: Learning-based Volume Rendering Time Prediction

A new deep learning model, ENTIRE, promises to enhance volume rendering time predictions, crucial for graphics programmers. By accurately forecasting rendering times based on various parameters, it allows for better …

Research arXiv cs.GR · 9 months, 2 weeks ago
Deep-BrownConrady: Prediction of Camera Calibration and Distortion Parameters Using Deep Learning and Synthetic Data

A new deep learning approach for camera calibration using synthetic data could significantly streamline workflows for graphics programmers. By leveraging a comprehensive dataset, this method allows for accurate parameter prediction …

Research arXiv cs.GR · 9 months, 2 weeks ago
Attention-guided reference point shifting for Gaussian-mixture-based partial point set registration

The introduction of an attention-based reference point shifting layer could significantly enhance partial point set registration in computer vision. This advancement is particularly relevant for graphics programmers, as it improves …

Research arXiv cs.GR · 9 months, 2 weeks ago