Projects sharing features with Multinerf

ml-sharp is a neural radiance field framework designed for single-image 3D reconstruction. It uses a neural network to predict 3D geometry and appearance from a single photograph in a single feedforward pass. The system generates metric 3D scene representations and includes a real-time view synthesizer for producing high-resolution images of new viewpoints. It also features a camera trajectory renderer that creates video sequences by moving a virtual camera through the predicted 3D space. The project covers coordinate-based neural rendering, 3D Gaussian representation regression, and real-ti

Threestudio is a 3D generative AI framework designed to create three-dimensional assets from text prompts and images. It provides specialized pipelines for text-to-3D generation and image-to-3D reconstruction, utilizing a neural radiance field trainer to produce geometry and textures. The framework is distinguished by its support for hybrid geometry backends, including signed distance functions, tetrahedra grids, and volume grids. It employs score distillation sampling to guide the generation process and features a modular plugin system for loading custom modules and nodes. The system covers

openMVS is a multi-view stereo library and photogrammetry pipeline used for 3D scene reconstruction. It transforms Structure from Motion data—specifically camera poses and sparse point clouds—into detailed 3D models consisting of dense point clouds and textured meshes. The project provides a sequence of processing stages to densify point clouds, generate 3D surface meshes, and apply photorealistic textures. It uses multi-view texture blending to map accurate colors onto reconstructed geometry and employs iterative refinement to optimize mesh details. The system includes capabilities for impo

Nerfstudio is a modular development framework for training, visualizing, and exporting three-dimensional scene representations derived from two-dimensional image datasets. It provides a neural scene reconstruction pipeline that converts raw images and camera data into high-fidelity 3D assets and cinematic video using a differentiable volumetric renderer. The system features an interactive web-based visualizer that allows users to monitor training progress and inspect neural scene geometry in real time. It decouples neural network architectures from the training loop through a standardized mod

This project is a computer vision pipeline and volumetric rendering system used to transform photos and videos into high-fidelity 3D models. It implements a deformable neural radiance field framework that optimizes deformation fields to represent non-rigid moving subjects in three dimensions. The system utilizes volumetric deformation fields to map 3D coordinates from a static canonical space to a deformed state. This allows for the reconstruction of photorealistic scenes and the synthesis of high-fidelity images from camera perspectives not present in the original input data. The framework

openMVG is a computer vision geometry library and toolkit for multiple view geometry. It serves as a framework for structure from motion and 3D scene reconstruction, providing the tools necessary to recover 3D point clouds and camera poses from collections of 2D images. The library implements both global and incremental structure-from-motion pipelines. It uses geometric algorithms to calculate camera pose estimation and image localization, employing Levenberg-Marquardt bundle adjustment to refine 3D coordinates and camera parameters by minimizing reprojection error. The project covers a broa

4DGaussians is a research library and neural rendering engine designed for reconstructing and rendering dynamic three-dimensional scenes. It represents moving environments as a collection of Gaussian primitives that evolve in position and appearance over a temporal dimension. The framework utilizes neural deformation fields to predict spatial offsets and rotations for static point representations, simulating complex motion over time. It further employs temporal basis decomposition to encode motion trajectories into learned functions, compressing dynamic scene data while maintaining smooth tra

This project is a comprehensive instructional resource and course for building neural networks using PyTorch. It covers the fundamental building blocks of deep learning, including tensor manipulation, automatic differentiation, and the construction of modular neural network components. The repository serves as a technical guide for several specialized domains. It provides implementation details for computer vision tasks such as image classification, object detection, and semantic segmentation, as well as natural language processing workflows involving transformers, recurrent networks, and gen

Gaussian Splatting is a computational framework designed to transform sparse sets of two-dimensional photographs into photorealistic, interactive three-dimensional scene representations. The system functions as a reconstruction tool and rendering engine, enabling the conversion of image data into volumetric models that support novel view synthesis. The project represents scenes as a collection of anisotropic three-dimensional Gaussians, which store position, opacity, color, and covariance data. It distinguishes itself through a differentiable tile-based rasterization process that projects the

This project is an RGB-D image inpainting tool and framework for 3D photo reconstruction. It transforms single 2D images into 3D content by estimating monocular depth and synthesizing missing color and depth data to fill occluded regions. The system uses a layered depth image representation to manage scene boundaries and pixel connectivity. This allows for novel view synthesis, enabling the generation of videos that simulate motion parallax effects from different camera perspectives. The project covers a range of spatial modeling capabilities, including depth map estimation, disparity-based

gsplat is a high-performance differentiable rasterization engine for 3D Gaussian splatting, designed for real-time novel view synthesis from 2D images. It provides a complete pipeline for reconstructing 3D scenes by optimizing differentiable Gaussian representations, training models from COLMAP-processed captures or proprietary device files, and generating new viewpoints through a CUDA-accelerated rendering backend. The framework distinguishes itself through memory-optimized CUDA kernels that reduce training memory usage by up to 4x compared to standard implementations while matching publishe

OpenSfM is a computer vision library and structure-from-motion pipeline designed to reconstruct three-dimensional scenes and camera trajectories from overlapping images. It functions as a 3D reconstruction engine and photogrammetry toolkit, utilizing automated feature-based image matching and incremental bundle adjustment to derive spatial geometry. The system distinguishes itself as a geospatial alignment tool, integrating GPS and inertial sensor data to align reconstructed 3D models with real-world geographic coordinates. It employs a hybrid Python and C++ execution model to manage large-sc

COLMAP is a 3D scene reconstruction suite and C++ geometry library that implements a full structure-from-motion pipeline. It functions as a GPU-accelerated photogrammetry tool and multi-view stereo framework designed to produce dense 3D geometry and watertight meshes from collections of 2D images. The project distinguishes itself through hardware-accelerated feature extraction and a modular camera modeling system that supports perspective, fisheye, and equirectangular lens types. It employs vocabulary tree image retrieval to efficiently identify similar images in large datasets and provides P

Neuralangelo is a neural surface reconstruction framework that transforms two-dimensional image sequences and multi-view photography into high-fidelity 3D meshes. It implements a pipeline for training neural radiance fields to represent complex scenes as digital geometry. The project utilizes a signed distance function for surface representation and multi-resolution hash encoding to capture both coarse and fine geometric details. It employs differentiable volume rendering and gradient-based eikonal regularization to ensure the learned distance functions remain physically plausible. The syste

This project is a collection of neural network models and geometric tools designed for image feature matching, spatial alignment, and visual localization. It provides a pre-trained neural network model for identifying high-accuracy correspondences between sparse image features without requiring local training. The system utilizes a graph neural network matcher that employs attention mechanisms and message passing to learn spatial relationships between image feature points. It integrates a RANSAC camera pose estimator to filter feature matches and calculate the relative spatial transformation

InstantMesh is a neural 3D reconstruction tool and single-image 3D mesh generator. It utilizes a sparse-view large reconstruction model to convert a single two-dimensional image into a three-dimensional object mesh. The system functions as a textured 3D mesh exporter, saving generated objects with either vertex colors or full texture maps for use in external rendering software. The framework covers a range of capabilities including feed-forward geometry inference, single-image depth estimation, and neural radiance fields. It also supports differentiable mesh rendering and workflows for spars

This project is a diffusion-based 3D generator and image-to-3D reconstruction system. It translates natural language descriptions or two-dimensional images into three-dimensional assets using neural radiance fields and diffusion models. The system utilizes score-distillation sampling and diffusion-based guidance to refine 3D shapes without requiring 3D training data. It includes specialized tools for transforming neural representations into exportable meshes with texture and material data, as well as a pipeline for iterative optimization of geometry and textures. The project covers a broad r

pifuhd is a 3D human reconstruction framework that generates high-resolution 3D meshes of people from a single 2D image. It utilizes pixel-aligned implicit functions to map image pixels to 3D space, predicting surface occupancy and distance to create detailed geometry. The system includes a pipeline for creating digital human assets, moving from 2D image feature projection to the extraction of discrete triangular meshes. It features specialized tools for refining these models, including a post-processor that removes geometric artifacts by isolating the largest connected component of the mesh.

Instant-ngp is a high-performance neural graphics engine and toolkit designed for 3D reconstruction and the rendering of neural radiance fields. It provides an integrated framework for generating photorealistic volumetric representations from sets of two-dimensional images by optimizing continuous neural scene models. The project distinguishes itself through a focus on rapid training and real-time inference, achieved by mapping spatial coordinates into compact feature grids. By utilizing multiresolution hash encoding and fused processing kernels, the system minimizes computational overhead an

This project is a static educational website and comprehensive curriculum focused on computer vision and deep learning. It serves as a public repository of instructional materials, lecture notes, and technical guides specifically detailing convolutional neural networks and visual recognition. The site is developed using static-site generation to host course documentation and student project directories. It provides structured academic resources that guide learners through image classification, generative modeling, and the implementation of various neural network architectures. The curriculum

Depth-Anything-3 is a collection of core model implementations for depth prediction, multi-view geometry estimation, and RGB-D spatial pipelines. It includes a monocular depth estimation model for predicting depth maps from single images or video, and a 3D Gaussian splatting generator that predicts parameters to synthesize high-fidelity novel views of a scene. The project provides a multi-view geometry estimator for calculating spatially consistent depth and camera poses across synchronized visual inputs. It also functions as a visual SLAM enhancement tool designed to reduce drift and improve

This is a framework for training and sampling diffusion models to generate high-fidelity images, video, and 4D assets. It provides a modular environment for managing generative AI training pipelines, including the handling of datasets, noise sampling, and loss weighting to stabilize the creation of synthetic content. The project features a modular model configuration system that uses YAML-based assembly to define network submodules and conditioners. It also includes a dedicated toolset for AI image watermarking, allowing for the embedding and detection of invisible markers to verify the origi

This project is a physically based rendering system and ray tracing engine designed to generate photorealistic images. It operates as a spectral rendering system that records radiance across discretized wavelength buckets and functions as a volumetric path tracer to compute light scattering and absorption within participating media. The engine utilizes GPU acceleration to execute its rendering pipeline on parallel graphics hardware. It integrates real-world optical data, such as measured spectral power distributions and lens description files, to simulate the behavior of physical camera syste

GET3D is a generative 3D mesh model and rendering framework designed to synthesize high-quality textured shapes and tetrahedral meshes. It functions as an image-to-3D reconstructor and text-to-3D generator, utilizing a differentiable 3D renderer to produce realistic visual perspectives and material effects. The system enables the creation of 3D assets from single 2D images, point clouds, or descriptive text prompts. It features a latent space interpolator for creating smooth transitions between different 3D objects and supports the independent control of geometry and texture. The project cov

Blender is a professional 3D creation suite designed for modeling, animation, rendering, and video editing. It functions as an open-source 3D engine that provides a comprehensive framework for procedural geometry, physics simulation, and high-quality visual output. The platform is built upon a foundational architecture that utilizes data-block-based memory management and a dependency-graph-based evaluation system to handle complex scene transformations and geometry updates. The software distinguishes itself through a highly modular, node-based procedural architecture that allows users to cons

This project is a computer vision system for monocular depth estimation and 3D point cloud generation. It provides a supervised depth learning framework and a depth predictor capable of estimating spatial distance and disparity from single 2D images using pretrained neural networks. The system includes tools to transform 2D depth images into 3D point clouds via pixel coordinate backprojection and converts 3D point cloud data into 2D depth maps. It utilizes a training pipeline that supports model fine-tuning and hyperparameter optimization. The library covers broader capabilities in spatial a

Neural renderer is a differentiable rendering library for PyTorch that projects three-dimensional meshes into two-dimensional images while maintaining continuous mathematical gradients for backpropagation. The framework enables gradient-based inverse rendering, allowing optimization of input parameters such as camera pose, vertex positions, and texture maps by propagating pixel-level reconstruction errors backward to the source geometry. The architecture incorporates approximate rasterisation gradients that substitute discontinuous edge derivatives with heuristic approximations to facilitate