Welcome to ImagenHub’s documentation!

ImagenHub is a centralized framework to standardize the evaluation of conditional image generation models by curating unified datasets, building an inference library and a benchmark that align with real life applications.

• Project Page

• Arxiv Paper

• HuggingFace Datasets

• GitHub Code

• Imagen Museum

Overview

• Introduction
  • Why ImagenHub?
• Philosophy
  • Purity of Evaluation
  • Research Roadmap
  • Open Collaboration
• Model Zoo
  • Text-guided Image Generation Model
  • Mask-guided Image Editing Model
  • Text-guided Image Editing Model
  • Subject-driven Image Generation Model
  • Subject-driven Image Editing Model
  • Multi-concept Image Composition Model
  • Control-guided Image Generation Model
• Dataset Zoo
  • Text-guided Image Generation Benchmark Dataset
  • Mask-guided Image Editing Benchmark Dataset
  • Text-guided Image Editing Benchmark Dataset
  • Subject-driven Image Generation Benchmark Dataset
  • Subject-driven Image Editing Benchmark Dataset
  • Multi-concept Image Composition Benchmark Dataset
  • Control-guided Image Generation Benchmark Dataset

Guidelines

• Installation
  • Environment Management
  • Extra Setup for Text-guided Image Generation
  • Extra Setup for Subject-driven Image Editing
  • Extra Setup for Multi-concept Image Composition
  • Miscellaneous
• Quickstart
  • Benchmarking
  • Running Metrics
  • Running single model
• Modify default benchmark
  • You want to test with different built-in models with ImagenHub dataset
• Learn By Examples
  • Ways to instantiate a model
  • Inference of model
  • Getting autometrics
  • Running custom benchmark using the ImagenHub library