Text-to-3D & Image-to-3D & Mesh Exportation with NeRF + Diffusion.

[ICCV 2023] Single-Stage Diffusion NeRF

An implementation using NVIDIA Kaolin Wisp and Stable Diffusion that applies the Score Distillation Sampling to the task of Few-shot / single-view 3D object reconstruction.

This repository contains the implementation of the paper: "Deceptive-NeRF/3DGS: Diffusion-Generated Pseudo-Observations for High-Quality Sparse-View Reconstruction", ECCV 2024.

No description available

a diffusion model trained through a subset of the LAION-5B dataset, and uses a frozen CLIP ViT-L/14 text encoder to condition the model on text prompts. Infrastructure consists of an autoencoder (VAE), defined U-Net, and a text-encoder to efficiently produce results with reduced memory and computing power.

Diffusion-guided NeRF pipeline for high-quality 3D scene reconstruction from RGBD images

Formulated a 3D NeRF inspection pipeline based on HF-NeuS (SDF) & diffusion-based model for 3D reconstruction of surface & crack segmentation respectively, improved HF-NeuS with two additional losses to reduce training time.

Text‑guided NeRF generation with diffusion: synthesize 3D neural radiance fields from natural language prompts

An AI-powered project for reconstructing color images from LiDAR depth data using GANs, NeRFs, or diffusion models.

A collection of AI/ML projects covering reinforcement learning, multi-agent systems, NLP, core ML techniques (clustering, SVM, trees, CNNs, autoencoders, diffusion), 3D reconstruction models (NeRF, NeuS), personal information detection, and fact-checking pipelines.

No description available

Code for diffusion, transformer and nerfs from scratch in pytorch.

No description available

No description available

Hybrid PCN + NeRF pipeline for 3D scan inpainting, achieving 0.978 SSIM on real-world datasets.

A General NeRF and Diffusion Toolkit for KatUI

No description available

数字人、3dgs、2dgs、nerf、diffusion models、smpl

No description available

The primary goal of the project is to develop a system that allows for the editing of NeRF scenes based on natural language instructions. This system builds on the strengths of diffusion models and advanced segmentation techniques to achieve localized and precise modifications in 3D scenes.

Evidentia is an AI-driven platform for automated crime scene reconstruction. It generates 2D images from textual descriptions using Stable Diffusion and converts images into text with ViT and GPT-2. Additionally, it processes FIR reports using T5 and BART to create structured crime narratives.

NeRFusion: Towards Understanding Diffusion and NeRF in Spatial Computing (Work in progress)

This repository contains my implementations of PointNet, NeRF, and diffusion using Pytorch

Implementation of Gaussian Splatting and Diffusion-guided Image, Mesh and NeRF optimizations

HMMAF: Hierarchical Multi-Modal Asset Factory. Recursive creative pipeline combining LLMs, VLMs, Diffusion, and NeRF models to generate, refine, and semantically enrich multimedia assets with quality assurance and brand compliance.

Formulated a 3D NeRF inspection pipeline based on HF-NeuS (SDF) & diffusion-based model for 3D reconstruction of surface & crack segmentation respectively, improved HF-NeuS with two additional losses to reduce training time.

This project is a Text-to-3D Model Generation and Repository System that allows users to generate, store, and share 3D assets using natural language prompts. The system leverages advanced Generative AI (NeRF, Gaussian Splatting, Diffusion Models, or Mesh-based pipelines) to convert text descriptions into 3D objects.

Project Idea: Train a Nerf of my living room and wrapp the model to a camera positioning engine (game engine), then querry images based on user inputs. At the same time, laverage a generative model (stable diffusion) such that the model may halucinate once the bounadries of the recorded data of the room has been reached.