Update README.md

README.md

@@ -1,3 +1,19 @@
+---
+language: 
+- en
+- zh
+library_name: transformers
+tags:
+- image-restoration
+- diffusion
+- computer-vision
+- flux
+- pytorch
+license: other
+license_name: flux-1-dev
+license_link: https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/LICENSE.md
+---
+
 <div align="center">
 <h1>🎨 LucidFlux:<br/>Caption-Free Universal Image Restoration with a Large-Scale Diffusion Transformer</h1>
 
@@ -27,8 +43,71 @@ Let us know if this works!
 ---
 
 ## 🌟 What is LucidFlux?
+
+<!-- <div align="center">
+<img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/demo/demo2.png" alt="What is LucidFlux - Quick Prompt Demo" width="1200"/>
+<br>
+</div> -->
+
 LucidFlux is a framework designed to perform high-fidelity image restoration across a wide range of degradations without requiring textual captions. By combining a Flux-based DiT backbone with Light-weight Condition Module and SigLIP semantic alignment, LucidFlux enables caption-free guidance while preserving structural and semantic consistency, achieving superior restoration quality.
 
+<!-- ## 🚀 Quick Start
+
+### 🔧 Installation
+
+```bash
+# Clone the repository
+git clone https://github.com/ephemeral182/LucidFlux.git
+cd LucidFlux
+
+# Create conda environment
+conda create -n postercraft python=3.11
+conda activate postercraft
+
+# Install dependencies
+pip install -r requirements.txt
+
+``` -->
+
+<!-- ### 🚀 Quick Generation
+
+Generate high-quality aesthetic posters from your prompt with `BF16` precision:
+
+```bash
+python inference.py \
+  --prompt "Urban Canvas Street Art Expo poster with bold graffiti-style lettering and dynamic colorful splashes" \
+  --enable_recap \
+  --num_inference_steps 28 \
+  --guidance_scale 3.5 \
+  --seed 42 \
+  --pipeline_path "black-forest-labs/FLUX.1-dev" \
+  --custom_transformer_path "LucidFlux/LucidFlux-v1_RL" \
+  --qwen_model_path "Qwen/Qwen3-8B"
+```
+
+If you are running on a GPU with limited memory, you can use `inference_offload.py` to offload some components to the CPU:
+
+```bash
+python inference_offload.py \
+  --prompt "Urban Canvas Street Art Expo poster with bold graffiti-style lettering and dynamic colorful splashes" \
+  --enable_recap \
+  --num_inference_steps 28 \
+  --guidance_scale 3.5 \
+  --seed 42 \
+  --pipeline_path "black-forest-labs/FLUX.1-dev" \
+  --custom_transformer_path "LucidFlux/LucidFlux-v1_RL" \
+  --qwen_model_path "Qwen/Qwen3-8B"
+``` -->
+<!-- 
+### 💻 Gradio Web UI
+
+We provide a Gradio web UI for LucidFlux. 
+
+```bash
+python demo_gradio.py
+``` -->
+
+
 ## 📊 Performance Benchmarks
 
 <div align="center">
@@ -195,6 +274,10 @@ LucidFlux is a framework designed to perform high-fidelity image restoration acr
 </tbody>
 </table>
 
+
+
+<!-- <img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/user_study/hpc.png" alt="User Study Results" width="1200"/> -->
+
 </div>
 
 ---
@@ -247,24 +330,25 @@ LucidFlux is a framework designed to perform high-fidelity image restoration acr
     <td width="200"><b>LQ</b></td>
     <td width="200"><b>HYPIR</b></td>
     <td width="200"><b>Topaz</b></td>
+    <td width="200"><b>SeeDream 4.0</b></td>
     <td width="200"><b>Gemini-NanoBanana</b></td>
     <td width="200"><b>GPT-4o</b></td>
     <td width="200"><b>Ours</b></td>
 </tr>
-<tr align="center"><td colspan="6"><img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/commercial_comparison/commercial_061.jpg" width="1200"></td></tr>
-<tr align="center"><td colspan="6"><img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/commercial_comparison/commercial_094.jpg" width="1200"></td></tr>
-<tr align="center"><td colspan="6"><img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/commercial_comparison/commercial_205.jpg" width="1200"></td></tr>
-<tr align="center"><td colspan="6"><img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/commercial_comparison/commercial_209.jpg" width="1200"></td></tr>
+<tr align="center"><td colspan="7"><img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/commercial_comparison/commercial_061.jpg" width="1400"></td></tr>
+<tr align="center"><td colspan="7"><img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/commercial_comparison/commercial_094.jpg" width="1400"></td></tr>
+<tr align="center"><td colspan="7"><img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/commercial_comparison/commercial_205.jpg" width="1400"></td></tr>
+<tr align="center"><td colspan="7"><img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/commercial_comparison/commercial_209.jpg" width="1400"></td></tr>
 </table>
 
 <details>
 <summary>Show more examples</summary>
 
 <table>
-<tr align="center"><td colspan="6"><img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/commercial_comparison/commercial_062.jpg" width="1200"></td></tr>
-<tr align="center"><td colspan="6"><img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/commercial_comparison/commercial_160.jpg" width="1200"></td></tr>
-<tr align="center"><td colspan="6"><img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/commercial_comparison/commercial_111.jpg" width="1200"></td></tr>
-<tr align="center"><td colspan="6"><img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/commercial_comparison/commercial_123.jpg" width="1200"></td></tr>
+<tr align="center"><td colspan="7"><img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/commercial_comparison/commercial_062.jpg" width="1400"></td></tr>
+<tr align="center"><td colspan="7"><img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/commercial_comparison/commercial_160.jpg" width="1400"></td></tr>
+<tr align="center"><td colspan="7"><img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/commercial_comparison/commercial_111.jpg" width="1400"></td></tr>
+<tr align="center"><td colspan="7"><img src="https://raw.githubusercontent.com/W2GenAI-Lab/LucidFlux/main/images/commercial_comparison/commercial_123.jpg" width="1400"></td></tr>
 </table>
 
 </details>
@@ -310,46 +394,33 @@ pip install -r requirements.txt
 ```
 
 ### Inference
-- **Flux.1 dev** → [🤗 FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev)  
-  Then update the model path in the `configs` for `flux-dev` in `src/flux/util.py` to your local FLUX.1-dev model path.
 
-- **T5** → [🤗 T5](https://huggingface.co/XLabs-AI/xflux_text_encoders)  
-  Then update the T5 path in the function `load_t5` in `src/flux/util.py` to your local T5 path.
+Prepare models in 2 steps, then run a single command.
+
+1) Login to Hugging Face (required for gated FLUX.1-dev). Skip if already logged-in.
 
-- **CLIP** → [🤗 CLIP](https://huggingface.co/openai/clip-vit-large-patch14)  
-  Then update the CLIP path in the function `load_clip` in `src/flux/util.py` to your local CLIP path.
+```bash
+python -m tools.hf_login --token "$HF_TOKEN"
+```
 
-- **SigLIP** → [🤗 siglip2-so400m-patch16-512](https://huggingface.co/google/siglip2-so400m-patch16-512)  
-  Then set `siglip_ckpt` to the corresponding local path.
+2) Download required weights to fixed paths and export env vars
 
-- **SwinIR** → [🤗 SwinIR](https://huggingface.co/lxq007/DiffBIR/blob/main/general_swinir_v1.ckpt)  
-  Then set `swin_ir_ckpt` to the corresponding local path.
+```bash
+# FLUX.1-dev (flow+ae), SwinIR prior, T5, CLIP, SigLIP and LucidFlux checkpoint to ./weights
+python -m tools.download_weights --dest weights
 
-- **LucidFlux** → [🤗 LucidFlux](https://huggingface.co/W2GenAI/LucidFlux)  
-  Then set `checkpoint` to the corresponding local path.
+# Exports FLUX_DEV_FLOW/FLUX_DEV_AE to your shell
+source weights/env.sh
+```
 
-```bash 
-inference.sh
 
-result_dir=ouput_images_folder
-input_folder=input_images_folder
-checkpoint_path=path/to/lucidflux.pth
-swin_ir_ckpt=path/to/swinir.ckpt
-siglip_ckpt=path/to/siglip.ckpt
+Run inference (uses fixed relative paths):
 
-mkdir -p ${result_dir}
-echo "Processing checkpoint..."
-python inference.py \
-  --checkpoint ${checkpoint_path} \
-  --swinir_pretrained ${swin_ir_ckpt} \
-  --control_image ${input_folder} \
-  --siglip_ckpt ${siglip_ckpt} \
-  --prompt "restore this image into high-quality, clean, high-resolution result" \
-  --output_dir ${result_dir}/ \
-  --width 1024 --height 1024 --num_steps 50 \
+```bash
+bash inference.sh
 ```
 
-Finially ```bash inference.sh```. You can also obtain the results of LucidFlux on RealSR and RealLQ250 from Hugging Face: [**LucidFlux**](https://huggingface.co/W2GenAI/LucidFlux).
+You can also obtain results of LucidFlux on RealSR and RealLQ250 from Hugging Face: [**LucidFlux**](https://huggingface.co/W2GenAI/LucidFlux).
 
 ## 🪪 License
 
@@ -383,4 +454,4 @@ For any questions or inquiries, please reach out to us:
 </details>
 
 
-</div>
+</div>