File size: 1,934 Bytes
fcc5627 220de69 fcc5627 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 | # CephVIT
CephVIT is a deep learning application for automatic **cephalometric landmark detection** from cephalogram images.
## Requirements
CephViT requires the following Python packages:
```text
torch
torchvision
torchaudio
timm
gradio
opencv-python
numpy
pillow
albumentations
cryptography
huggingface_hub
pandas
pydantic
tqdm
```
All dependencies are included in `requirements.txt` and can be installed with:
```bash
pip install -r requirements.txt
```
## Run Locally
Clone the repository:
```bash
git clone https://huggingface.co/nlm-dir/CephViT
cd CephViT
```
Install dependencies:
```bash
pip install -r requirements.txt
```
The included model checkpoint is encrypted. Set the `MODEL_KEY` environment variable before running:
```bash
export MODEL_KEY="2d65697cf6ad50c95a0a2558c73658a7"
```
Then start the app:
```bash
python app.py
```
## Project Structure
```text
CephVIT/
├── app.py # Gradio inference application
├── model.py # Model definitions
├── heatmap_utils.py # Heatmap generation and decoding
├── secure_torch_load.py # Encrypted checkpoint loader
├── best.pt.enc # Encrypted model checkpoint
└── requirements.txt
```
## License
OpenMDW 1.1.
## Citation
If you use CephViT in your work, please cite:
```bibtex
@inproceedings{hou2026automatic,
title = {Automatic Cephalometric Landmark Localization on {CBCT}-Derived Digitally Reconstructed Radiographs for Skeletal Malocclusion Classification},
author = {Hou, Benjamin and Almpani, Konstantinia and Lee, Janice S. and Lu, Zhiyong},
booktitle = {Oral and Dental Image Analysis (ODIN 2026)},
series = {Lecture Notes in Computer Science},
publisher = {Springer},
year = {2026},
note = {Accepted for publication}
}
```
*Accepted to ODIN 2026. Citation details will be updated when the Springer proceedings are published.*
|