| --- |
| library_name: pytorch |
| base_model: facebook/VGGT-1B |
| datasets: |
| - czZhang2026/ViewRecDB-100K |
| tags: |
| - computer-vision |
| - viewpoint-recommendation |
| - ai-photography |
| --- |
| |
| <div align="center"> |
| <h1>ViewRecNet</h1> |
| <p><strong>ViewRec3D: Learning to Recommend 3D Viewpoints for AI Photography</strong></p> |
| <p> |
| <a href="https://github.com/anonymous1-submission/ViewRec3D/blob/main/ViewRec3D.pdf"><img src="https://img.shields.io/badge/Paper-ViewRec3D-blue" alt="Paper"></a> |
| <a href="https://github.com/anonymous1-submission/ViewRec3D"><img src="https://img.shields.io/badge/GitHub-Code-24292f" alt="Code"></a> |
| <a href="https://huggingface.co/datasets/czZhang2026/ViewRecDB-100K"><img src="https://img.shields.io/badge/Dataset-ViewRecDB--100K-orange" alt="Dataset"></a> |
| </p> |
| <p>Chengzhi Zhang<sup>1,*</sup>, Yunzhong Hou<sup>1,*,†</sup>, Zizheng Sun<sup>1</sup>, Ming-Hsuan Yang<sup>2</sup>, Chi Harold Liu<sup>1</sup></p> |
| <p><sup>1</sup> Beijing Institute of Technology <sup>2</sup> University of California, Merced</p> |
| <p><sup>*</sup> Equal contribution. <sup>†</sup> Corresponding author.</p> |
| </div> |
| |
| ## Overview |
| |
| Given a single photograph, ViewRecNet recommends how to move and rotate the camera, adjust the field of view, and choose portrait or landscape orientation to approach an expert composition. The model uses a VGGT backbone and learns from paired views in ViewRecDB-100K with dense camera-ray supervision. |
| |
| ## Released checkpoint |
| |
| `model.safetensors` (4.51 GB) contains the weights from training step 15,625, including the backbone, camera head, orientation head, and ray head. The evaluation code constructs the network and loads these weights directly. |
| |
| ## Quick start |
| |
| From the ViewRec3D code repository, install the project requirements and download the checkpoint: |
| |
| ```bash |
| python -m pip install -r requirements.txt |
| python -m pip install -U huggingface_hub |
| hf download czZhang2026/ViewRecNet model.safetensors --local-dir checkpoints/ViewRecNet |
| ``` |
| |
| To evaluate on a CUDA GPU, prepare the test split as described in the [dataset card](https://huggingface.co/datasets/czZhang2026/ViewRecDB-100K). Replace the test directory below with its local path. The evaluator reads the checkpoint from `--output_dir` and writes evaluation metrics there. |
| |
| ```bash |
| python -m src.evaluate --val_dir /path/to/ViewRecDB-100K/test --output_dir checkpoints/ViewRecNet --use_rays True |
| ``` |
| |
| ## Acknowledgements |
| |
| This work was supported by the National Natural Science Foundation of China (No. U25A20434). |
| |
