Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,238 @@
+<div align="center">
+<h1> SubjectGenius </h1>
+
+<h3>SubjectGenius: Unified Multi-Conditional Combination with Diffusion Transformer</h3>
+<b>Haoxuan Wang</b>, Jinlong Peng, Qingdong He, Hao Yang, Ying Jin, <br>
+ Jiafu Wu, Xiaobin Hu, Yanjie Pan, Zhenye Gan, Mingmin Chi, Bo Peng, Yabiao Wang <br>
+<br>
+<a href="https://arxiv.org/abs/2503.09277"><img src="https://img.shields.io/badge/arXiv-2503.09277-A42C25.svg" alt="arXiv"></a>
+<a href="https://huggingface.co/Xuan-World/SubjectGenius"><img src="https://img.shields.io/badge/🤗_HuggingFace-Model-ffbd45.svg" alt="HuggingFace"></a>
+<a href="https://huggingface.co/datasets/Xuan-World/SubjectSpatial200K"><img src="https://img.shields.io/badge/🤗_HuggingFace-Dataset-ffbd45.svg" alt="HuggingFace"></a>
+</div>
+
+## 🌠 Key Features
+
+<img src='assets/cover.png' width='100%' />
+<br>
+Fantastic results of our proposed SubjectGenius on multi-conditional controllable generation: <br>
+
+- (a) Subject-Insertion task. 
+- (b) and (c) Subject-Spatial task. 
+- (d) Multi-Spatial task.
+
+Our unified framework effectively handles any combination of input conditions and achieves remarkable alignment with all of them, including but not limited to text prompts, spatial maps, and subject images.
+
+## 🚩 **Updates**
+
+- ✅ March 12, 2025. We release SubjectSpatial200K dataset.
+- ✅ March 12, 2025. We release SubjectGenius framework.
+
+## 🔧 Dependencies and Installation
+
+```bash
+conda create -n SubjectGenius python=3.12
+conda activate SubjectGenius
+pip install -r requirements.txt
+```
+Due to the issues of _diffusers_ library, you need to update the `cite-package` code manually.
+You can find the location of your _diffusers_ library by running the following command.
+```bash
+pip show diffusers
+```
+
+Then add the following entry to the dictionary `_SET_ADAPTER_SCALE_FN_MAPPING` located in `diffusers/loaders/peft.py`:
+```
+"SubjectGeniusTransformer2DModel": lambda model_cls, weights: weights
+```
+
+## 📥 Download Models 
+Place all the model weights in the `ckpt` directory. Of course, it's also acceptable to store them in other directories.
+1. **FLUX.1-schnell**
+```bash
+huggingface-cli download black-forest-labs/FLUX.1-schnell --local-dir ./ckpt/FLUX.1-schnell
+```
+2. **Condition-LoRA**
+```bash
+huggingface-cli download Xuan-World/SubjectGenius --include "Condition_LoRA/*" --local-dir ./ckpt/Condition_LoRA
+```
+
+3. **Denoising-LoRA**
+```bash
+huggingface-cli download Xuan-World/SubjectGenius --include "Denoising_LoRA/*" --local-dir ./ckpt/Denoising_LoRA
+```
+
+4. FLUX.1-schnell-training-assistant-LoRA (optional) 
+
+Download it if you want to train your LoRA on the FLUX-schnell.
+
+```bash
+huggingface-cli download ostris/FLUX.1-schnell-training-adapter --local-dir ./ckpt/FLUX.1-schnell-training-adapter
+```
+
+> Schnell is a step distilled model, meaning it can generate an image in just a few steps. 
+> However, this makes it impossible to train on it directly because every step you train breaks down the compression more and more. 
+> With this adapter enabled during training, that doesnt happen. 
+> It is activated during the training process, and disabled during sampling. 
+> After the LoRA is trained, this adapter is no longer needed.
+
+## 🎮 Inference on Demo
+- We provide the `inference.py` script to offer a simplest and fastest way for you to run our model. <br>
+- Replace the arguments `--version` from `training-based` to `training-free`, then you don't need to provide the **Denoising-LoRA** module.
+- Adjust the scale of `--denoising_lora_weight` to get a balance between the editability and the consistency when using Custom Prompts.
+### 1. Subject-Insertion
+Default Prompts：
+```bash
+python inference.py \
+--condition_types fill subject \
+--denoising_lora ckpt/Denoising_LoRA/subject_fill_union \
+--denoising_lora_weight 1.0 \
+--fill examples/window/background.jpg \
+--subject examples/window/subject.jpg \
+--json "examples/window/1634_rank0_A decorative fabric topper for windows..json" \
+--version training-based
+```
+
+### 2. Subject-Canny
+Default Prompts：
+```bash
+python inference.py \
+--condition_types canny subject \
+--denoising_lora ckpt/Denoising_LoRA/subject_canny_union \
+--denoising_lora_weight 1.0 \
+--canny examples/doll/canny.jpg \
+--subject examples/doll/subject.jpg \
+--json "examples/doll/1116_rank0_A spooky themed gothic doll..json" \
+--version training-based
+```
+Custom Prompts：
+```bash
+python inference.py \
+--condition_types canny subject \
+--denoising_lora ckpt/Denoising_LoRA/subject_canny_union \
+--denoising_lora_weight 0.6 \
+--canny examples/doll/canny.jpg \
+--subject examples/doll/subject.jpg \
+--json "examples/doll/1116_rank0_A spooky themed gothic doll..json" \
+--version training-based \
+--prompt "She stands amidst the vibrant glow of a bustling Chinatown alley, \
+her pink hair shimmering under festive lantern light, clad in a sleek black dress adorned with intricate lace patterns. "
+```
+### 3. Subject-Depth
+Default Prompts：
+```bash
+python inference.py \
+--condition_types depth subject \
+--denoising_lora ckpt/Denoising_LoRA/subject_depth_union \
+--denoising_lora_weight 1.0 \
+--depth examples/car/depth.jpg \
+--subject examples/car/subject.jpg \
+--json "examples/car/2532_rank0_A sturdy ATV with rugged looks..json" \
+--version training-based
+```
+Custom Prompts：
+```bash
+python inference.py \
+--condition_types depth subject \
+--denoising_lora ckpt/Denoising_LoRA/subject_depth_union \
+--denoising_lora_weight 0.6 \
+--depth examples/car/depth.jpg \
+--subject examples/car/subject.jpg \
+--json "examples/car/2532_rank0_A sturdy ATV with rugged looks..json" \
+--version training-based \
+--prompt "It is positioned on a snow-covered path in a forest, its green body dusted with frost and black tires caked with packed snow. \
+The vehicle retains its sturdy build with handlebars glinting ice particles and headlights cutting through falling snowflakes, surrounded by tall pine trees draped in white."
+```
+### 4. Depth-Canny
+Default Prompts：
+```bash
+python inference.py \
+--condition_types depth canny \
+--denoising_lora ckpt/Denoising_LoRA/depth_canny_union \
+--denoising_lora_weight 1.0 \
+--depth examples/toy/depth.jpg \
+--canny examples/toy/canny.jpg \
+--json "examples/toy/1616_rank0_A soft, plush toy with cuddly features..json" \
+--version training-based
+```
+Custom Prompts：
+```bash
+python inference.py \
+--condition_types depth canny \
+--denoising_lora ckpt/Denoising_LoRA/depth_canny_union \
+--denoising_lora_weight 0.6 \
+--depth examples/toy/depth.jpg \
+--canny examples/toy/canny.jpg \
+--json "examples/toy/1616_rank0_A soft, plush toy with cuddly features..json" \
+--version training-based \
+--prompt "It sits on a moonlit sandy beach, a small sandcastle partially washed by gentle tides beside it, \
+under a night sky where the full moon casts silvery trails across waves, with distant seagulls gliding through star-dappled darkness."
+```
+
+## 🗂️ Download Dataset （optional）
+1. Download SubjectSpatial200K
+
+Place our SubjectSpatial200K dataset in the `dataset` directory. Of course, it's also acceptable to store them in other directories. <br>
+
+```bash
+huggingface-cli download Xuan-World/SubjectSpatial200K --repo-type dataset --local-dir ./dataset
+```
+
+2. Filter and Partition the SubjectSpatial200K dataset into training and testing sets.
+
+The default partition scheme is identical to our paper.
+You can customize it as you wish.
+
+```bash
+python src/partition_dataset.py \
+--dataset dataset/SubjectSpatial200K/data_labeled \
+--output_dir dataset/split_SubjectSpatial200K \
+--partition train
+```
+```bash
+python src/partition_dataset.py \
+--dataset dataset/SubjectSpatial200K/Collection3/data_labeled \
+--output_dir dataset/split_SubjectSpatial200K/Collection3 \
+--partition train
+```
+```bash
+python src/partition_dataset.py \
+--dataset dataset/SubjectSpatial200K/data_labeled \
+--output_dir dataset/split_SubjectSpatial200K \
+--partition test
+```
+```bash
+python src/partition_dataset.py \
+--dataset dataset/SubjectSpatial200K/Collection3/data_labeled \
+--output_dir dataset/split_SubjectSpatial200K/Collection3 \
+--partition test
+```
+## 🧩 Train in single-conditional setting
+Refer to https://github.com/Yuanshi9815/OminiControl to train your **Condition-LoRA** modules.  We will release our reimplementation using diffusers soon.
+
+## 🔥 Train in multi-conditional setting
+Use our SubjectSpatial200K dataset or your customized multi-conditional dataset to train your **Denoising-LoRA** module. 
+1. Configure Accelerate Environment
+```bash
+accelerate config
+```
+2. Launch Distributed Training
+```bash
+accelerate launch train.py
+```
+
+## 📊 Batch Inference on Dataset
+- We provide a script for batch inference on the SubjectSpatial200K dataset in both training-free and training-based version.
+- It can also be run on your custom datasets through your Dataset and DataLoader implementations.
+```bash
+python test.py
+```
+
+## 📚 Citation
+```
+@article{wang2025SubjectGenius,
+  title={SubjectGenius: Unified Multi-Conditional Combination with Diffusion Transformer},
+  author={Wang, Haoxuan and Peng, Jinlong and He, Qingdong and Yang, Hao and Jin, Ying and Wu, Jiafu and Hu, Xiaobin and Pan, Yanjie and Gan, Zhenye and Chi, Mingmin and others},
+  journal={arXiv preprint arXiv:2503.09277},
+  year={2025}
+}
+```