Duplicate from TyangJN/FIG

.gitattributes

@@ -0,0 +1,59 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.lz4 filter=lfs diff=lfs merge=lfs -text
+*.mds filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+# Audio files - uncompressed
+*.pcm filter=lfs diff=lfs merge=lfs -text
+*.sam filter=lfs diff=lfs merge=lfs -text
+*.raw filter=lfs diff=lfs merge=lfs -text
+# Audio files - compressed
+*.aac filter=lfs diff=lfs merge=lfs -text
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+*.ogg filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text
+# Image files - uncompressed
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+# Image files - compressed
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+# Video files - compressed
+*.mp4 filter=lfs diff=lfs merge=lfs -text
+*.webm filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,231 @@
+---
+license: apache-2.0
+configs:
+  - config_name: Prompts
+    data_files:
+    - split: prompt
+      path: prompts.jsonl
+  - config_name: Reference
+    data_files:
+    - split: Animal
+      path: txt/animal/animal_reference.jsonl
+    - split: Culture
+      path: txt/culture/culture_reference.jsonl
+    - split: Event
+      path: txt/event/event_reference.jsonl
+    - split: Food
+      path: txt/food/food_reference.jsonl
+    - split: Landmarks
+      path: txt/landmarks/landmarks_reference.jsonl
+    - split: People
+      path: txt/people/people_reference.jsonl
+    - split: Plant
+      path: txt/plant/plant_reference.jsonl
+    - split: Product
+      path: txt/product/product_reference.jsonl
+    - split: Sports
+      path: txt/sports/sports_reference.jsonl
+    - split: Transportation
+      path: txt/transportation/transportation_reference.jsonl
+  - config_name: Evaluation
+    data_files:
+    - split: Animal
+      path: txt/animal/animal_evaluation.jsonl
+    - split: Culture
+      path: txt/culture/culture_evaluation.jsonl
+    - split: Event
+      path: txt/event/event_evaluation.jsonl
+    - split: Food
+      path: txt/food/food_evaluation.jsonl
+    - split: Landmarks
+      path: txt/landmarks/landmarks_evaluation.jsonl
+    - split: People
+      path: txt/people/people_evaluation.jsonl
+    - split: Plant
+      path: txt/plant/plant_evaluation.jsonl
+    - split: Product
+      path: txt/product/product_evaluation.jsonl
+    - split: Sports
+      path: txt/sports/sports_evaluation.jsonl
+    - split: transportation
+      path: txt/transportation/transportation_evaluation.jsonl
+  - config_name: Images
+    data_files:
+    - split: Reference
+      path: img/img_reference.jsonl
+language:
+- en
+- zh
+tags:
+- agent
+size_categories:
+- 1K<n<10K
+---
+
+# ORIG: Multi-Modal Retrieval-Enhanced Image Generation
+[![arXiv](https://img.shields.io/badge/arXiv-Paper-b31b1b.svg)](https://www.arxiv.org/abs/2510.22521)
+[![GitHub](https://img.shields.io/badge/GitHub-Repository-black)](https://github.com/TyangJN/ORIG)
+[![Project](https://img.shields.io/badge/Project-Website-green)](https://tyangjn.github.io/orig.github.io/)
+
+Large Multimodal Models (LMMs) have achieved remarkable progress in generating photorealistic and prompt-aligned images, but they often produce outputs that contradict verifiable knowledge, especially when prompts involve fine-grained attributes or time-sensitive events. Conventional retrieval-augmented approaches attempt to address this issue by introducing external information, yet they are fundamentally incapable of grounding generation in accurate and evolving knowledge due to their reliance on static sources and shallow evidence integration. To bridge this gap, we introduce ORIG, an agentic open multimodal retrieval-augmented framework for Factual Image Generation (FIG), a new task that requires both visual realism and factual grounding. ORIG iteratively retrieves and filters multimodal evidence from the web and incrementally integrates the refined knowledge into enriched prompts to guide generation. To support systematic evaluation, we build FIG-Eval, a benchmark spanning ten categories across perceptual, compositional, and temporal dimensions. Experiments demonstrate that ORIG substantially improves factual consistency and overall image quality over strong baselines, highlighting the potential of open multimodal retrieval for factual image generation.
+
+## 🎯 Key Features
+
+- **🌐 Open Web Retrieval**: Dynamically retrieves up-to-date information from the web
+- **🔄 Iterative Refinement**: Multi-round evidence gathering and prompt enhancement
+- **📊 Multimodal Integration**: Combines text and image evidence for comprehensive grounding
+- **🎭 Multi-Modal Support**: GPT and Qwen retrieval models with various generation backends
+- **📈 Comprehensive Evaluation**: FIG-Eval benchmark across 10 categories
+
+## 🚀 Quick Start
+
+### Installation
+
+```bash
+pip install torch transformers openai pillow requests
+pip install google-generativeai serpapi qwen-vl-utils
+```
+
+### Basic Usage
+
+```python
+from orig import ORIGPipeline
+
+# Initialize the pipeline
+pipeline = ORIGPipeline(
+    retrieval_model="gpt",  # or "qwen"
+    generation_model="openai_gen",  # or "gemini_gen", "qwen_gen"
+    max_rounds=3
+)
+
+# Generate factually grounded image
+result = pipeline.generate(
+    prompt="Generate a photo of frog lifecycle",
+    modality="mm"  # multimodal retrieval
+)
+
+print(f"Generated image: {result['image_path']}")
+print(f"Retrieval summary: {result['retrieval_summary']}")
+```
+
+### API Configuration
+
+```python
+import os
+
+# Set your API keys
+os.environ["OPENAI_API_KEY"] = "your_openai_key"
+os.environ["GOOGLE_API_KEY"] = "your_google_key"
+os.environ["SERPAPI_API_KEY"] = "your_serpapi_key"
+```
+
+## 📊 Supported Modalities
+
+| Modality | Description | Use Case |
+|----------|-------------|----------|
+| `mm` | Multimodal retrieval | Comprehensive factual grounding |
+| `txt` | Text-only retrieval | Factual descriptions and context |
+| `img` | Image-only retrieval | Visual style and appearance |
+| `cot` | Chain-of-thought | Step-by-step reasoning |
+| `dir` | Direct generation | Baseline without retrieval |
+
+
+## 🔧 Advanced Configuration
+
+### Custom Retrieval Rounds
+
+```python
+pipeline = ORIGPipeline(
+    retrieval_model="gpt",
+    generation_model="openai_gen",
+    max_rounds=5,  # More rounds for complex queries
+    batch_size=10
+)
+```
+
+### Local Model Support
+
+```python
+# Use local Qwen model
+pipeline = ORIGPipeline(
+    retrieval_model="qwen_local",
+    model_path="Qwen/Qwen2.5-VL-7B-Instruct",
+    device="cuda"
+)
+```
+
+## 📚 Dataset: FIG-Eval
+
+The FIG-Eval benchmark includes 10 categories:
+
+- **Animal**: Lifecycle, behavior, habitat
+- **Culture**: Traditions, artifacts, ceremonies  
+- **Event**: Historical moments, celebrations
+- **Food**: Traditional dishes, preparation
+- **Landmarks**: Famous structures, locations
+- **People**: Historical figures, professions
+- **Plant**: Growth stages, species
+- **Product**: Design evolution, usage
+- **Sports**: Rules, equipment, techniques
+- **Transportation**: Vehicle evolution, operation
+
+### Download Dataset
+
+```python
+from datasets import load_dataset
+
+dataset = load_dataset("TyangJN/FIG-Eval")
+```
+
+## 🔍 Retrieval Process
+
+1. **Warm-up**: Identify key entities and concepts
+2. **Iterative Search**: Multi-round evidence gathering
+3. **Evidence Filtering**: Quality assessment and filtering
+4. **Prompt Enhancement**: Integration into generation prompts
+5. **Image Generation**: Factually grounded output
+
+## 🛠️ Model Details
+
+### Retrieval Models
+- **GPT-4**: OpenAI's multimodal model for text and image understanding
+- **Qwen2.5-VL**: Alibaba's vision-language model for comprehensive analysis
+
+### Generation Models
+- **OpenAI GPT-Image-1**
+- **Google Nano Banana**
+- **Qwen Qwen-Image**
+
+## 📖 Citation
+
+```bibtex
+@article{orig2024,
+  title={ORIG: Multi-Modal Retrieval-Enhanced Image Generation Evaluation System},
+  author={},
+  journal={},
+  year={2025}
+}
+```
+
+## 🤝 Contributing
+
+We welcome contributions! Please see our [GitHub repository](https://github.com/your-username/ORIG) for:
+- Issue reporting
+- Feature requests  
+- Pull requests
+- Documentation improvements
+
+## 📄 License
+
+This project is licensed under the MIT License. See the [LICENSE](https://github.com/your-username/ORIG/blob/main/LICENSE) file for details.
+
+## 🔗 Links
+
+- **📄 Paper**: [arXiv:2510.22521](https://arxiv.org/abs/2401.00000)
+- **🌐 Demo**: [Web Project](https://tyangjn.github.io/orig.github.io/)
+- **📊 Dataset**: [FIG-Eval](https://huggingface.co/datasets/TyangJN/FIG/viewer)
+- **💻 Code**: [GitHub Repository](https://github.com/TyangJN/ORIG)
+
+---
+
+**Note**: This model requires API keys for external services. Please configure your keys before use.