Papers
Papers
Embedded Files
Neural Two-Level Monte Carlo Real-Time Rendering
Neural Two-Level Monte Carlo Real-Time Rendering
Mikhail Dereviannykh, Dmitrii Klepikov, Johannes Hanika, Carsten Dachsbacher
Eurographics 2025, Computer Graphics Forum
⭐Honorable Mention⭐ from the "Best Paper Award Committee"
TLTR: We introduce Neural Incident Radiance Cache (NIRC) enabling efficient integration over a neural network domain instead of relying on costly path tracing. To ensure unbiased estimations and improved accuracy, we employ a Two-Level Monte Carlo approach, outperforming Neural Control Variates in rendering quality.
Arxiv Version | Video Demo (High Res, Low Res)
Arxiv Version | Video Demo (High Res, Low Res)
Real-Time Path-Guiding Based on Parametric Mixture Model
Real-Time Path-Guiding Based on Parametric Mixture Model
Mikhail Derevyannykh
Eurographics 2022
Path-Guiding algorithms for sampling scattering directions can drastically decrease the variance of Monte Carlo estimators of Light Transport Equation, but their usage was limited to offline rendering because of memory and computational limitations. We introduce a new robust screen-space technique that is based on online learning of parametric mixture models for guiding the real-time path-tracing algorithm. It requires storing of 8 parameters for every pixel, achieves a reduction of FLIP metric up to 4 times with 1 spp rendering. Also, it consumes less than 1.5ms on RTX 2070 for 1080p and reduces path-tracing timings by generating more coherent rays by about 5% on average. Moreover, it leads to significant bias reduction and a lower level of flickering of SVGF output.
Conference Version | Fast Forward | Video Presentation | Video Demo
Stohastic Spherical Gaussians Radiance Volumes For
Stohastic Spherical Gaussians Radiance Volumes For
Production Real-time Rendering
Production Real-time Rendering
Bachelor Thesis, 2021
Mikhail Derevyannykh
We present the results of a study of the fast, flexible, and robust solution of the global illumination problem based on precomputed light volumes with stochastically filtered Spherical Gaussians (SG). It supports coherent shading of dynamic and static objects, taking into account visibility weighting for preventing light and shadow leaks. In addition, our lighting basis and optimization algorithm of its parameters are robust for high-frequency indirect lighting and achieve lower errors concerning the current state-of-the-art methods based on SG. Furthermore, we suggest a more accurate approach for computing Fresnel, Smith Masking Function, and Cos factors of light transport equation when incident lighting is approximated by SG basis. And finally, a method for estimating local high-frequency occlusion with SG basis is presented, too.
ECS - Data Oriented Ray Tracer
ECS - Data Oriented Ray Tracer
Course Work, 2019
Mikhail Derevyannykh
Wooden PBR Engine is a software unidirectional path-tracing researching engine for rendering 3D scenes based on C++17, SIMD Math Library, Data-Oriented ECS Design (similar to Unity ECS). It has some very good low-level performance metrics with regard to OOP RayTracers (PBRT) - some research plots and comparisons are inside the paper. I implemented my own SIMD math library, ECS meta-programming library, allocators library for this project.
Previous not published research
Previous not published research
Lightmap and light volumes predictions for statis scenes using CNNs
Lightmap and light volumes predictions for statis scenes using CNNs
Models predict a lightmap/light-volume given dynamic lights input in a feed-forward way for a real-time game. 3rd-year university project
Pipeline based on Blender+Colab+Weight and Biases+Google Drive+Ngrok+TensorRT for researching and finding the best neural network architecture.
Weight and Biases for logging and hyperparameter searching (architecture types, data transformation, # of layers, learning rates, # samples for implemented custom loss functions, activation functions, etc)
Deployed render-server based on Ngrok for fast rendering validation screenshots and for benchmarking inference time of requested neural network architecture on some gaming GPU.
Renderer with support for highly optimized async neural network inference and dataset generation
Sophisticated data generation and post-processing. Terabytes of data, optimized RAM\Disk memory access, and workload. Efficient GPU-Kernels for some data-preprocessing blocks.
Light field is described in Spherical Gaussians basis (The Order 1886) with DDGI-Driven Probe-Placement Tricks
Inference is about 1 ms, 2060 RTX
Page updated
Google Sites
Report abuse