Official implementation of "RaPD: Resolution-Agnostic Pixel Diffusion via Semantics-Enriched Implicit Representations" (NeurIPS 2026).
-
Updated
Nov 20, 2026 - Python
Official implementation of "RaPD: Resolution-Agnostic Pixel Diffusion via Semantics-Enriched Implicit Representations" (NeurIPS 2026).
DDPM implemented from scratch in PyTorch: generates handwritten digits from pure noise. FID 3.60 (real-data reference 2.20), no mode collapse, no memorization.
从 CNN 到 3DGS:一个月 AI 基础训练手册(6 个模块 · 30 个 Demo · 8 篇笔记)
ComfyUI custom node for AI image generation with Krea 2, Z-Image, Anima, Qwen Image, Qwen Image 2.1 (Turbo) and MOSS-Image. Optional PixelDiT (PiD) 2x/4x upscaling, multi-LoRA and Korean/English UI.
This repository provides step-by-step instructions for setting up ComfyUI and integrating Stable Diffusion models for AI image generation. It includes guidance on downloading, extracting, and correctly placing model files to ensure smooth operation within ComfyUI.
LungSynth AI is an advanced medical web application designed to synthesize intermediate 4D lung CT scan phases from two boundary scans. Featuring a modern UI, the platform allows researchers to upload scans, run AI inference, and visualize dynamic respiratory motion.
Benchmark and comparison dashboard for FLUX.1-dev spatial acceleration methods (Baseline vs. TaylorSeer vs. RALU) built with Streamlit.
Python pipeline for automated garment detail correction with ComfyUI
An open, hands-on deep learning course covering neural networks, backpropagation, CNNs, Transformers, LLMs, generative AI, diffusion models, GNNs, and reinforcement learning through interactive demos, mathematical explanations, PyTorch labs, and practical exercises.
Reproducibility study of BUDDI: Generative Proxemics for 3D human pose estimation
A STATS 100A Honors study of class-conditional diffusion for Chinese character images, with a saved checkpoint and an interactive local demo.
Code and measurement tables: a controlled study of preservation-mask sources in training-free video editing
LiveSeed: score every seed of a text-to-image diffusion model from attention entropy in the first denoising steps, and finish only the best. Code, human-study data, prompts and seeds.
To associate your repository with the diffusion-models-for-image-generation topic, visit your repo's landing page and select "manage topics."