Views
No views yet
Despite recent advances in personalized image generation, existing models consistently fail to produce reliable multi-human scenes, often merging or losing facial identity. We present Ar2Can, a novel two-stage framework that disentangles spatial planning from identity rendering for multi-human generation. The Architect predicts structured layouts, specifying where each person should appear. The Artist then synthesizes photorealistic images, guided by a spatially-grounded face matching reward that combines Hungarian spatial alignment with identity similarity. This approach ensures faces are rendered at correct locations and faithfully preserve reference identities. We develop two Architect variants, seamlessly integrated with our diffusion-based Artist model. This is optimized via Group Relative Policy Optimization (GRPO) using compositional rewards for count accuracy, image quality, and identity matching. Evaluated on the MultiHuman-Testbench, Ar2Can achieves substantial improvements in both count accuracy and identity preservation, while maintaining high perceptual quality. Notably, our method achieves these results using primarily synthetic data, without requiring real multi-human images.
| Adapter | Subdirectory | Base model | Purpose |
|---|---|---|---|
| Architect | architect/ | black-forest-labs/FLUX.1-schnell | Generates scene layout / probe image |
| Artist | artist/ | black-forest-labs/FLUX.1-Kontext-dev | Renders final identity-preserving composition |
| Field | Value |
|---|---|
| Adapter type | LoRA (PEFT) |
| LoRA rank (r) | 64 |
| LoRA alpha | 128 |
| Dropout | 0.0 |
| Weight init | Gaussian |
attn.to_q, attn.to_k, attn.to_v, attn.to_out.0, attn.to_add_out, attn.add_q_proj, attn.add_k_proj, attn.add_v_proj, ff.net.0.proj, ff.net.2, ff_context.net.0.proj, ff_context.net.2attn.to_q, attn.to_k, attn.to_v, attn.to_out.0, attn.to_add_out, attn.add_q_proj, attn.add_k_proj, attn.add_v_proj, ff.net.0.proj, ff.net.2, ff_context.net.0.proj, ff_context.net.2, proj_mlp1git clone https://github.com/Qualcomm-AI-research/ar2can.git
2cd ar2can1python infer.py \
2 --faces faces/person_1.webp faces/person_2.webp faces/person_3.webp \
3 --prompt "Three people standing side-by-side in a winter landscape, 8K, ultra-realistic, preserve exact facial identity."1python app.py
2# open http://localhost:7865loras/architect/adapter_model.safetensors and loras/artist/adapter_model.safetensors — the code will use the local copies automatically.Two people on a sunny beach, ultra-realistic, 8K, preserve exact facial identity.
Three people standing side-by-side in a winter landscape, realistic light rendering, preserve exact facial identity.
Four people in a city plaza at midday, high fidelity, preserve exact facial identity.1@inproceedings{borse2026ar2can,
2 title={Ar2can: An architect and an artist leveraging a canvas for multi-human generation},
3 author={Borse, Shubhankar and Pham, Phuc and Farhadzadeh, Farzad and Choi, Seokeon and Nguyen, Phong and Tran, Anh and Yun, Sungrack and Hayat, Munawar and Porikli, Fatih},
4 booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
5 pages={550--560},
6 year={2026}
7}Disclaimer: The model weights released here are not identical to the model reported in the paper. They are provided for research and demonstration purposes only. Performance may vary from the results reported in the paper.